* Sync with the trunk.

when they are defined

Makefile.am

@@ -2,6 +2,8 @@ SUBDIRS = externals src scripts corepkgs doc misc tests
 EXTRA_DIST = substitute.mk nix.spec nix.spec.in bootstrap.sh \
   nix.conf.example NEWS version
 
+pkginclude_HEADERS = config.h
+
 include ./substitute.mk
 
 nix.spec: nix.spec.in

aterm-gc.supp

@@ -115,3 +115,35 @@
    fun:*
    fun:AT_collect
 }
+
+{
+   ATerm library conservatively scans for GC roots
+   Memcheck:Value4
+   fun:*
+   fun:*
+   fun:mark_phase
+}
+
+{
+   ATerm library conservatively scans for GC roots
+   Memcheck:Cond
+   fun:*
+   fun:*
+   fun:mark_phase
+}
+
+{
+   ATerm library conservatively scans for GC roots
+   Memcheck:Value4
+   fun:*
+   fun:*
+   fun:mark_phase_young
+}
+
+{
+   ATerm library conservatively scans for GC roots
+   Memcheck:Cond
+   fun:*
+   fun:*
+   fun:mark_phase_young
+}

bootstrap.sh

@@ -1,4 +1,5 @@
 #! /bin/sh -e
+rm -f aclocal.m4
 mkdir -p config
 libtoolize --copy
 aclocal

configure.ac

@@ -10,8 +10,8 @@ AC_CANONICAL_HOST
 
 # Construct a Nix system name (like "i686-linux").
 AC_MSG_CHECKING([for the canonical Nix system name])
-cpu_name=$(uname -p | tr 'A-Z ' 'a-z_')
-machine_name=$(uname -m | tr 'A-Z ' 'a-z_')
+cpu_name=$(uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ ' 'abcdefghijklmnopqrstuvwxyz_')
+machine_name=$(uname -m | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ ' 'abcdefghijklmnopqrstuvwxyz_')
 
 case $machine_name in
     i*86)
@@ -30,7 +30,7 @@ case $machine_name in
         ;;
 esac
 
-sys_name=$(uname -s | tr 'A-Z ' 'a-z_')
+sys_name=$(uname -s | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ ' 'abcdefghijklmnopqrstuvwxyz_')
 
 case $sys_name in
     cygwin*)
@@ -50,48 +50,49 @@ AC_DEFINE_UNQUOTED(SYSTEM, ["$system"], [platform identifier (`cpu-os')])
 test "$localstatedir" = '${prefix}/var' && localstatedir=/nix/var
 
 
-# Whether to produce a statically linked binary.  On Cygwin, this is
-# the default: dynamically linking against the ATerm DLL does work,
-# except that it requires the ATerm "lib" directory to be in $PATH, as
-# Windows doesn't have anything like an RPATH embedded in executable.
-# Since this is kind of annoying, we use static libraries for now.
-
-AC_ARG_ENABLE(static-nix, AC_HELP_STRING([--enable-static-nix],
-  [produce statically linked binaries]),
-  static_nix=$enableval, static_nix=no)
-
-if test "$sys_name" = cygwin; then
-   static_nix=yes
-fi
-
-if test "$static_nix" = yes; then
+# Windows-specific stuff.  On Cygwin, dynamically linking against the
+# ATerm DLL works, except that it requires the ATerm "lib" directory
+# to be in $PATH, as Windows doesn't have anything like an RPATH
+# embedded in executable.  Since this is kind of annoying, we use
+# static libraries for now.
+if test "$sys_name" = "cygwin"; then
     AC_DISABLE_SHARED
     AC_ENABLE_STATIC
 fi
 
 
-# Windows-specific stuff.
-if test "$sys_name" = "cygwin"; then
-    # We cannot delete open files.
-    AC_DEFINE(CANNOT_DELETE_OPEN_FILES, 1, [Whether it is impossible to delete open files.])
-fi
-
 # Solaris-specific stuff.
 if test "$sys_name" = "sunos"; then
     # Solaris requires -lsocket -lnsl for network functions
-    ADDITIONAL_NETWORK_LIBS="-lsocket -lnsl"
-    AC_SUBST(ADDITIONAL_NETWORK_LIBS)
+    LIBS="-lsocket -lnsl $LIBS"
 fi
 
+
 AC_PROG_CC
 AC_PROG_CXX
 
+# To build programs to be run in the build machine
+if test "$CC_FOR_BUILD" = ""; then
+    if test "$cross_compiling" = "yes"; then
+        AC_CHECK_PROGS(CC_FOR_BUILD, gcc cc)
+    else
+        CC_FOR_BUILD="$CC"
+    fi
+fi
+AC_SUBST([CC_FOR_BUILD])
 
 # We are going to use libtool.
 AC_DISABLE_STATIC
 AC_ENABLE_SHARED
 AC_PROG_LIBTOOL
 
+if test "$enable_shared" = yes; then
+   SUB_CONFIGURE_FLAGS="--enable-shared --disable-static"
+else
+   SUB_CONFIGURE_FLAGS="--enable-static --disable-shared"
+fi
+AC_SUBST(SUB_CONFIGURE_FLAGS)
+
 
 # Use 64-bit file system calls so that we can support files > 2 GiB.
 AC_SYS_LARGEFILE
@@ -127,11 +128,22 @@ AC_CHECK_HEADERS([locale], [], [], [])
 AC_LANG_POP(C++)
 
 
+# Check for <err.h>.
+AC_CHECK_HEADER([err.h], [], [bsddiff_compat_include="-Icompat-include"])
+AC_SUBST([bsddiff_compat_include])
+
+
 # Check whether we have the personality() syscall, which allows us to
 # do i686-linux builds on x86_64-linux machines.
 AC_CHECK_HEADERS([sys/personality.h])
 
 
+# Check for tr1/unordered_set.
+AC_LANG_PUSH(C++)
+AC_CHECK_HEADERS([tr1/unordered_set], [], [], [])
+AC_LANG_POP(C++)
+
+
 AC_DEFUN([NEED_PROG],
 [
 AC_PATH_PROG($1, $2)
@@ -149,6 +161,7 @@ AC_PATH_PROG(w3m, w3m, false)
 AC_PATH_PROG(flex, flex, false)
 AC_PATH_PROG(bison, bison, false)
 NEED_PROG(perl, perl)
+NEED_PROG(sed, sed)
 NEED_PROG(tar, tar)
 AC_PATH_PROG(dot, dot)
 AC_PATH_PROG(dblatex, dblatex)
@@ -193,23 +206,6 @@ AC_ARG_WITH(store-dir, AC_HELP_STRING([--with-store-dir=PATH],
   storedir=$withval, storedir='/nix/store')
 AC_SUBST(storedir)
 
-AC_ARG_WITH(aterm, AC_HELP_STRING([--with-aterm=PATH],
-  [prefix of CWI ATerm library]),
-  aterm=$withval, aterm=)
-AM_CONDITIONAL(HAVE_ATERM, test -n "$aterm")
-if test -z "$aterm"; then
-  aterm_lib='-L${top_builddir}/externals/inst-aterm/lib -lATerm'
-  aterm_include='-I${top_builddir}/externals/inst-aterm/include'
-  aterm_bin='${top_builddir}/externals/inst-aterm/bin'
-else
-  aterm_lib="-L$aterm/lib -lATerm"
-  aterm_include="-I$aterm/include"
-  aterm_bin="$aterm/bin"
-fi
-AC_SUBST(aterm_lib)
-AC_SUBST(aterm_include)
-AC_SUBST(aterm_bin)
-
 AC_ARG_WITH(openssl, AC_HELP_STRING([--with-openssl=PATH],
   [prefix of the OpenSSL library]),
   openssl=$withval, openssl=)
@@ -225,6 +221,8 @@ AC_ARG_WITH(bzip2, AC_HELP_STRING([--with-bzip2=PATH],
   [prefix of bzip2]),
   bzip2=$withval, bzip2=)
 AM_CONDITIONAL(HAVE_BZIP2, test -n "$bzip2")
+ATERM_VERSION=2.5
+AC_SUBST(ATERM_VERSION)
 if test -z "$bzip2"; then
   # Headers and libraries will be used from the temporary installation
   # in externals/inst-bzip2.
@@ -245,8 +243,36 @@ AC_SUBST(bzip2_include)
 AC_SUBST(bzip2_bin)
 AC_SUBST(bzip2_bin_test)
 
+AC_ARG_WITH(sqlite, AC_HELP_STRING([--with-sqlite=PATH],
+  [prefix of SQLite]),
+  sqlite=$withval, sqlite=)
+AM_CONDITIONAL(HAVE_SQLITE, test -n "$sqlite")
+SQLITE_VERSION=3070500
+AC_SUBST(SQLITE_VERSION)
+if test -z "$sqlite"; then
+  sqlite_lib='${top_builddir}/externals/sqlite-autoconf-$(SQLITE_VERSION)/libsqlite3.la'
+  sqlite_include='-I${top_builddir}/externals/sqlite-autoconf-$(SQLITE_VERSION)'
+  sqlite_bin='${top_builddir}/externals/sqlite-autoconf-$(SQLITE_VERSION)'
+else
+  sqlite_lib="-L$sqlite/lib -lsqlite3"
+  sqlite_include="-I$sqlite/include"
+  sqlite_bin="$sqlite/bin"
+fi
+AC_SUBST(sqlite_lib)
+AC_SUBST(sqlite_include)
+AC_SUBST(sqlite_bin)
 
-AC_CHECK_LIB(pthread, pthread_mutex_init)
+# Whether to use the Boehm garbage collector.
+AC_ARG_ENABLE(gc, AC_HELP_STRING([--enable-gc],
+  [enable garbage collection in the Nix expression evaluator (requires Boehm GC)]),
+  gc=$enableval, gc=)
+if test -n "$gc"; then
+  PKG_CHECK_MODULES([BDW_GC], [bdw-gc])
+  boehmgc_lib="-L$boehmgc/lib -lgc"
+  CXXFLAGS="$BDW_GC_CFLAGS $CXXFLAGS"
+  AC_DEFINE(HAVE_BOEHMGC, 1, [Whether to use the Boehm garbage collector.])
+fi
+AC_SUBST(boehmgc_lib)
 
 
 AC_ARG_ENABLE(init-state, AC_HELP_STRING([--disable-init-state],
@@ -260,9 +286,7 @@ AC_CHECK_FUNCS([setresuid setreuid lchown])
 
 
 # Nice to have, but not essential.
-AC_CHECK_FUNCS([strsignal])
-AC_CHECK_FUNCS([posix_fallocate])
-AC_CHECK_FUNCS([fdatasync])
+AC_CHECK_FUNCS([strsignal posix_fallocate nanosleep])
 
 
 # This is needed if ATerm or bzip2 are static libraries,
@@ -272,14 +296,6 @@ if test "$(uname)" = "Darwin"; then
 fi
 
 
-if test "$static_nix" = yes; then
-    # `-all-static' has to be added at the end of configure, because
-    # the C compiler doesn't know about -all-static (it's filtered out
-    # by libtool, but configure doesn't use libtool).
-    LDFLAGS="-all-static $LDFLAGS"
-fi
-
-
 AM_CONFIG_HEADER([config.h])
 AC_CONFIG_FILES([Makefile
    externals/Makefile

corepkgs/buildenv/builder.pl.in

@@ -29,10 +29,18 @@ sub createLinks {
         $baseName =~ s/^.*\///g; # strip directory
         my $dstFile = "$dstDir/$baseName";
 
+        # The files below are special-cased so that they don't show up
+        # in user profiles, either because they are useless, or
+        # because they would cause pointless collisions (e.g., each
+        # Python package brings its own
+        # `$out/lib/pythonX.Y/site-packages/easy-install.pth'.)
         # Urgh, hacky...
-	if ($srcFile =~ /\/propagated-build-inputs$/ ||
+        if ($srcFile =~ /\/propagated-build-inputs$/ ||
             $srcFile =~ /\/nix-support$/ ||
             $srcFile =~ /\/perllocal.pod$/ ||
+            $srcFile =~ /\/easy-install.pth$/ ||
+            $srcFile =~ /\/site.py$/ ||
+            $srcFile =~ /\/site.pyc$/ ||
             $srcFile =~ /\/info\/dir$/ ||
             $srcFile =~ /\/log$/)
         {
@@ -160,4 +168,4 @@ while (scalar(keys %postponed) > 0) {
 print STDERR "created $symlinks symlinks in user environment\n";
 
 
-symlink($ENV{"manifest"}, "$out/manifest") or die "cannot create manifest";
+symlink($ENV{"manifest"}, "$out/manifest.nix") or die "cannot create manifest";

corepkgs/buildenv/default.nix

@@ -11,4 +11,8 @@ derivation {
   paths = derivations;
   active = map (x: if x ? meta && x.meta ? active then x.meta.active else "true") derivations;
   priority = map (x: if x ? meta && x.meta ? priority then x.meta.priority else "5") derivations;
+
+  # Building user environments remotely just causes huge amounts of
+  # network traffic, so don't do that.
+  preferLocalBuild = true;
 }

corepkgs/nar/nar.sh.in

@@ -7,8 +7,6 @@ dst=$out/tmp.nar.bz2
 
 @bzip2@ < tmp > $dst
 
-@bindir@/nix-hash -vvvvv --flat --type $hashAlgo --base32 tmp > $out/nar-hash
-
 @bindir@/nix-hash --flat --type $hashAlgo --base32 $dst > $out/narbz2-hash
 
 @coreutils@/mv $out/tmp.nar.bz2 $out/$(@coreutils@/cat $out/narbz2-hash).nar.bz2

doc/manual/conf-file.xml

@@ -97,6 +97,25 @@ env-keep-derivations = false
   </varlistentry>
 
 
+  <varlistentry xml:id="conf-build-cores"><term><literal>build-cores</literal></term>
+
+    <listitem><para>Sets the value of the
+    <envar>NIX_BUILD_CORES</envar> environment variable in the
+    invocation of builders.  Builders can use this variable at their
+    discretion to control the maximum amount of parallelism.  For
+    instance, in Nixpkgs, if the derivation attribute
+    <varname>enableParallelBuilding</varname> is set to
+    <literal>true</literal>, the builder passes the
+    <option>-j<replaceable>N</replaceable></option> flag to GNU Make.
+    It can be overriden using the <option
+    linkend='opt-cores'>--cores</option> command line switch and
+    defaults to <literal>1</literal>.  The value <literal>0</literal>
+    means that the builder should use all available CPU cores in the
+    system.</para></listitem>
+
+  </varlistentry>
+
+
   <varlistentry xml:id="conf-build-max-silent-time"><term><literal>build-max-silent-time</literal></term>
 
     <listitem>
@@ -241,7 +260,7 @@ build-use-chroot = /dev /proc /bin</programlisting>
     Nix store metadata (in <filename>/nix/var/nix/db</filename>) are
     synchronously flushed to disk.  This improves robustness in case
     of system crashes, but reduces performance.  The default is
-    <literal>false</literal>.</para></listitem>
+    <literal>true</literal>.</para></listitem>
 
   </varlistentry>
     

doc/manual/env-common.xml

@@ -271,6 +271,17 @@ $ mount -o bind /mnt/otherdisk/nix /nix</screen>
 
 </varlistentry>
 
+
+<varlistentry><term><envar>GC_INITIAL_HEAP_SIZE</envar></term>
+
+  <listitem><para>If Nix has been configured to use the Boehm garbage
+  collector, this variable sets the initial size of the heap in bytes.
+  It defaults to 384 MiB.  Setting it to a low value reduces memory
+  consumption, but will increase runtime due to the overhead of
+  garbage collection.</para></listitem>
+
+</varlistentry>
+
     
 </variablelist>
 

doc/manual/installation.xml

@@ -96,17 +96,22 @@ ubiquitous 2.5.4a won't.  Note that these are only required if you
 modify the parser or when you are building from the Subversion
 repository.</para>
 
-<para>Nix uses CWI's ATerm library and the bzip2 compressor (including
-the bzip2 library).  These are included in the Nix source
-distribution.  If you build from the Subversion repository, you must
-download them yourself and place them in the
-<filename>externals/</filename> directory.  See
+<para>Nix uses the bzip2 compressor (including the bzip2 library).  It
+is included in the Nix source distribution.  If you build from the
+Subversion repository, you must download it yourself and place it in
+the <filename>externals/</filename> directory.  See
 <filename>externals/Makefile.am</filename> for the precise URLs of
-these packages.  Alternatively, if you already have them installed,
-you can use <command>configure</command>'s
-<option>--with-aterm</option> and <option>--with-bzip2</option>
+this packages.  Alternatively, if you already have it installed, you
+can use <command>configure</command>'s <option>--with-bzip2</option>
 options to point to their respective locations.</para>
 
+<para>Nix can optionally use the <link
+xlink:href="http://www.hpl.hp.com/personal/Hans_Boehm/gc/">Boehm
+garbage collector</link> to reduce the evaluator’s memory consumption.
+To enable it, install <literal>pkgconfig</literal> and the Boehm
+garbage collector, and pass the flag <option>--enable-gc</option> to
+<command>configure</command>.</para>
+
 </section>
 
 

doc/manual/introduction.xml

@@ -113,7 +113,7 @@ $ nix-env --rollback
 
 <simplesect><title>Garbage collection</title>
 
-<para>When you install a package like this…
+<para>When you uninstall a package like this…
 
 <screen>
 $ nix-env --uninstall firefox

doc/manual/manual.xml

@@ -27,10 +27,11 @@
       <year>2007</year>
       <year>2008</year>
       <year>2009</year>
+      <year>2010</year>
       <holder>Eelco Dolstra</holder>
     </copyright>
 
-    <date>September 2009</date>
+    <date>August 2010</date>
     
   </info>
 

doc/manual/nix-push.xml

@@ -120,7 +120,7 @@ dependencies used in the build, such as compilers).</para>
 dependencies, we can do:
       
 <screen>
-$ nix-push <replaceable>urls</replaceable> $(nix-instantiate $(nix-store -r foo.nix))</screen>
+$ nix-push <replaceable>urls</replaceable> $(nix-store -r $(nix-instantiate foo.nix))</screen>
     
 </para>
 

doc/manual/nix-store.xml

@@ -342,7 +342,7 @@ $ nix-store --gc --max-freed $((100 * 1024 * 1024))</screen>
 
 <cmdsynopsis>
   <command>nix-store</command>
-  <arg choice='plain'><option>--gc</option></arg>
+  <arg choice='plain'><option>--delete</option></arg>
   <arg><option>--ignore-liveness</option></arg>
   <arg choice='plain' rep='repeat'><replaceable>paths</replaceable></arg>
 </cmdsynopsis>
@@ -404,6 +404,7 @@ error: cannot delete path `/nix/store/zq0h41l75vlb4z45kzgjjmsjxvcv1qk7-mesa-6.4'
     <arg choice='plain'><option>--tree</option></arg>
     <arg choice='plain'><option>--binding</option> <replaceable>name</replaceable></arg>
     <arg choice='plain'><option>--hash</option></arg>
+    <arg choice='plain'><option>--size</option></arg>
     <arg choice='plain'><option>--roots</option></arg>
   </group>
   <arg><option>--use-output</option></arg>
@@ -587,9 +588,21 @@ query is applied to the target of the symlink.</para>
   <varlistentry><term><option>--hash</option></term>
   
     <listitem><para>Prints the SHA-256 hash of the contents of the
-    store paths <replaceable>paths</replaceable>.  Since the hash is
-    stored in the Nix database, this is a fast
-    operation.</para></listitem>
+    store paths <replaceable>paths</replaceable> (that is, the hash of
+    the output of <command>nix-store --dump</command> on the given
+    paths).  Since the hash is stored in the Nix database, this is a
+    fast operation.</para></listitem>
+
+  </varlistentry>
+
+  <varlistentry><term><option>--size</option></term>
+  
+    <listitem><para>Prints the size in bytes of the contents of the
+    store paths <replaceable>paths</replaceable> — to be precise, the
+    size of the output of <command>nix-store --dump</command> on the
+    given paths.  Note that the actual disk space required by the
+    store paths may be higher, especially on filesystems with large
+    cluster sizes.</para></listitem>
 
   </varlistentry>
 

doc/manual/opt-common-syn.xml

@@ -13,6 +13,10 @@
   </group>
   <replaceable>number</replaceable>
 </arg>
+<arg>
+  <arg><option>--cores</option></arg>
+  <replaceable>number</replaceable>
+</arg>
 <arg>
   <arg><option>--max-silent-time</option></arg>
   <replaceable>number</replaceable>

doc/manual/opt-common.xml

@@ -98,7 +98,25 @@
   linkend='conf-build-max-jobs'><literal>build-max-jobs</literal></link>
   configuration setting, which itself defaults to
   <literal>1</literal>.  A higher value is useful on SMP systems or to
-  exploit I/O latency.  </para></listitem>
+  exploit I/O latency.</para></listitem>
+  
+</varlistentry>
+
+
+<varlistentry xml:id="opt-cores"><term><option>--cores</option></term>
+
+  <listitem><para>Sets the value of the <envar>NIX_BUILD_CORES</envar>
+  environment variable in the invocation of builders.  Builders can
+  use this variable at their discretion to control the maximum amount
+  of parallelism.  For instance, in Nixpkgs, if the derivation
+  attribute <varname>enableParallelBuilding</varname> is set to
+  <literal>true</literal>, the builder passes the
+  <option>-j<replaceable>N</replaceable></option> flag to GNU Make.
+  It defaults to the value of the <link
+  linkend='conf-build-cores'><literal>build-cores</literal></link>
+  configuration setting, if set, or <literal>1</literal> otherwise.
+  The value <literal>0</literal> means that the builder should use all
+  available CPU cores in the system.</para></listitem>
   
 </varlistentry>
 

doc/manual/package-management.xml

@@ -38,7 +38,7 @@ to end-user applications like Mozilla Firefox.  (Nix is however not
 tied to the Nix Package collection; you could write your own Nix
 expressions based on it, or completely new ones.)  You can download
 the latest version from <link
-xlink:href='http://nixos.org/releases/full-index-nixpkgs.html' />.</para>
+xlink:href='http://nixos.org/nixpkgs/download.html' />.</para>
 
 <para>Assuming that you have downloaded and unpacked a release of Nix
 Packages, you can view the set of available packages in the release:

doc/manual/quick-start.xml

@@ -60,7 +60,7 @@ available remotely.</para></listitem>
 in the channel:
 
 <screen>
-$ nix-env -qa ’*’ <lineannotation>(mind the quotes!)</lineannotation>
+$ nix-env -qa \*
 docbook-xml-4.2
 firefox-1.0pre-PR-0.10.1
 hello-2.1.1

doc/manual/release-notes.xml

@@ -6,6 +6,92 @@
 
 
 
+<!--==================================================================-->
+
+<section xml:id="ssec-relnotes-1.0"><title>Release 1.0 (TBA)</title>
+
+<para>This release has the following improvements:</para>
+
+<itemizedlist>
+
+  <listitem>
+    <para>Nix can now optionally use the Boehm garbage collector.
+    This significantly reduces the Nix evaluator’s memory footprint,
+    especially when evaluating large NixOS system configurations.  It
+    can be enabled using the <option>--enable-gc</option> configure
+    option.</para>
+  </listitem>
+
+</itemizedlist>
+
+</section>
+
+
+<!--==================================================================-->
+
+<section xml:id="ssec-relnotes-0.16"><title>Release 0.16 (August 17, 2010)</title>
+
+<para>This release has the following improvements:</para>
+
+<itemizedlist>
+
+  <listitem>
+    <para>The Nix expression evaluator is now much faster in most
+    cases: typically, <link
+    xlink:href="http://www.mail-archive.com/nix-dev@cs.uu.nl/msg04113.html">3
+    to 8 times compared to the old implementation</link>.  It also
+    uses less memory.  It no longer depends on the ATerm
+    library.</para>
+  </listitem>
+
+  <listitem>
+    <para>
+      Support for configurable parallelism inside builders.  Build
+      scripts have always had the ability to perform multiple build
+      actions in parallel (for instance, by running <command>make -j
+      2</command>), but this was not desirable because the number of
+      actions to be performed in parallel was not configurable.  Nix
+      now has an option <option>--cores
+      <replaceable>N</replaceable></option> as well as a configuration
+      setting <varname>build-cores =
+      <replaceable>N</replaceable></varname> that causes the
+      environment variable <envar>NIX_BUILD_CORES</envar> to be set to
+      <replaceable>N</replaceable> when the builder is invoked.  The
+      builder can use this at its discretion to perform a parallel
+      build, e.g., by calling <command>make -j
+      <replaceable>N</replaceable></command>.  In Nixpkgs, this can be
+      enabled on a per-package basis by setting the derivation
+      attribute <varname>enableParallelBuilding</varname> to
+      <literal>true</literal>.
+    </para>
+  </listitem>
+
+  <listitem>
+    <para><command>nix-store -q</command> now supports XML output
+    through the <option>--xml</option> flag.</para>
+  </listitem>
+
+  <listitem>
+    <para>Several bug fixes.</para>
+  </listitem>
+
+</itemizedlist>
+
+</section>
+
+
+<!--==================================================================-->
+
+<section xml:id="ssec-relnotes-0.15"><title>Release 0.15 (March 17, 2010)</title>
+
+<para>This is a bug-fix release.  Among other things, it fixes
+building on Mac OS X (Snow Leopard), and improves the contents of
+<filename>/etc/passwd</filename> and <filename>/etc/group</filename>
+in <literal>chroot</literal> builds.</para>
+
+</section>
+
+
 <!--==================================================================-->
 
 <section xml:id="ssec-relnotes-0.14"><title>Release 0.14 (February 4, 2010)</title>

externals/Makefile.am

@@ -1,35 +1,3 @@
-# CWI ATerm
-
-ATERM = aterm-2.4.2-fixes-r2
-
-$(ATERM).tar.bz2:
-	@echo "Nix requires the CWI ATerm library to build."
-	@echo "Please download version 2.4.2-fixes-r2 from"
-	@echo "  http://nixos.org/tarballs/aterm-2.4.2-fixes-r2.tar.bz2"
-	@echo "and place it in the externals/ directory."
-	false
-
-$(ATERM): $(ATERM).tar.bz2
-	bunzip2 < $(srcdir)/$(ATERM).tar.bz2 | tar xvf -
-
-have-aterm:
-	$(MAKE) $(ATERM)
-	touch have-aterm
-
-if HAVE_ATERM
-build-aterm:
-else
-build-aterm: have-aterm
-	(pfx=`pwd` && \
-	cd $(ATERM) && \
-	CC="$(CC)" ./configure --prefix=$$pfx/inst-aterm \
-	  --disable-shared --enable-static && \
-	$(MAKE) && \
-	$(MAKE) install)
-	touch build-aterm
-endif
-
-
 # bzip2
 
 BZIP2 = bzip2-1.0.5
@@ -44,30 +12,56 @@ $(BZIP2).tar.gz:
 $(BZIP2): $(BZIP2).tar.gz
 	gunzip < $(srcdir)/$(BZIP2).tar.gz | tar xvf -
 
-have-bzip2:
-	$(MAKE) $(BZIP2)
-	touch have-bzip2
-
 if HAVE_BZIP2
 build-bzip2:
 else
-build-bzip2: have-bzip2
-	(pfx=`pwd` && \
-	cd $(BZIP2) && \
-	$(MAKE) && \
-	$(MAKE) install PREFIX=$$pfx/inst-bzip2)
+build-bzip2: $(BZIP2)
+	(cd $(BZIP2) && \
+	$(MAKE) CC="$(CC)" && \
+	$(MAKE) install PREFIX=$(abs_builddir)/inst-bzip2)
 	touch build-bzip2
 
-install:
+install-exec-local:: build-bzip2
 	mkdir -p $(DESTDIR)${bzip2_bin}
 	$(INSTALL_PROGRAM) $(bzip2_bin_test)/bzip2 $(bzip2_bin_test)/bunzip2 $(DESTDIR)${bzip2_bin}
 endif
 
 
-all: build-aterm build-bzip2
+# SQLite
 
-EXTRA_DIST = $(ATERM).tar.bz2 $(BZIP2).tar.gz
+SQLITE = sqlite-autoconf-$(SQLITE_VERSION)
+SQLITE_TAR = sqlite-autoconf-$(SQLITE_VERSION).tar.gz
 
-ext-clean:
-	$(RM) -f have-aterm build-aterm have-bzip2 build-bzip2
-	$(RM) -rf $(ATERM) $(BZIP2)
+$(SQLITE_TAR):
+	@echo "Nix requires the SQLite library to build."
+	@echo "Please download version $(SQLITE_VERSION) from"
+	@echo "  http://www.sqlite.org/$(SQLITE_TAR)"
+	@echo "and place it in the externals/ directory."
+	false
+
+$(SQLITE): $(SQLITE_TAR)
+	gzip -d < $(srcdir)/$(SQLITE_TAR) | tar xvf -
+
+if HAVE_SQLITE
+build-sqlite:
+else
+build-sqlite: $(SQLITE)
+	(cd $(SQLITE) && \
+	CC="$(CC)" CFLAGS="-DSQLITE_ENABLE_COLUMN_METADATA=1" ./configure --disable-static --prefix=$(pkglibdir)/dummy --libdir=${pkglibdir} $(SUB_CONFIGURE_FLAGS) && \
+	$(MAKE) )
+	touch build-sqlite
+
+install-exec-local:: build-sqlite
+	cd $(SQLITE) && $(MAKE) install
+	rm -rf "$(DESTDIR)/$(pkglibdir)/dummy"
+endif
+
+
+all: build-bzip2 build-sqlite
+
+EXTRA_DIST = $(BZIP2).tar.gz $(SQLITE_TAR)
+
+clean:
+	$(RM) -f build-bzip2 build-sqlite
+	$(RM) -rf $(BZIP2) $(SQLITE)
+	$(RM) -rf inst-bzip2

nix.conf.example

@@ -59,6 +59,17 @@
 #build-max-jobs = 1
 
 
+### Option `build-cores'
+#
+# This option defines the number of CPU cores to utilize in parallel
+# within a build job, i.e. by passing an appropriate `-jN' flag to GNU
+# Make. The default is 1, meaning that parallel building within jobs
+# is disabled.  Passing the special value `0' causes Nix to try and
+# auto-detect the number of available cores on the local host.  This
+# setting can be overridden using the `--cores' command line switch.
+#build-cores = 1
+
+
 ### Option `build-max-silent-time'
 #
 # This option defines the maximum number of seconds that a builder can

release.nix

@@ -1,4 +1,7 @@
-{ nixpkgs ? ../nixpkgs }:
+{ nixpkgs ? ../nixpkgs
+, nix ? { outPath = ./.; rev = 1234; }
+, officialRelease ? false
+}:
 
 let
 
@@ -6,10 +9,6 @@ let
 
 
     tarball =
-      { nix ? {outPath = ./.; rev = 1234;}
-      , officialRelease ? false
-      }:
-
       with import nixpkgs {};
 
       releaseTools.sourceTarball {
@@ -19,8 +18,8 @@ let
         inherit officialRelease;
 
         buildInputs =
-          [ curl bison flex2533 perl libxml2 libxslt w3m bzip2
-            tetex dblatex nukeReferences
+          [ curl bison24 flex2535 perl libxml2 libxslt w3m bzip2
+            tetex dblatex nukeReferences pkgconfig
           ];
 
         configureFlags = ''
@@ -29,14 +28,14 @@ let
           --with-xml-flags=--nonet
         '';
 
-        # Include the ATerm and Bzip2 tarballs in the distribution.
+        # Include the Bzip2 tarball in the distribution.
         preConfigure = ''
-          stripHash ${aterm242fixes.src}
-          cp -pv ${aterm242fixes.src} externals/$strippedName
-
           stripHash ${bzip2.src}
           cp -pv ${bzip2.src} externals/$strippedName
 
+          stripHash ${sqlite.src}
+          cp -pv ${sqlite.src} externals/$strippedName
+
           # TeX needs a writable font cache.
           export VARTEXFONTS=$TMPDIR/texfonts
         '';
@@ -63,70 +62,43 @@ let
 
 
     build =
-      { tarball ? jobs.tarball {}
-      , system ? "i686-linux"
-      }:
+      { system ? "i686-linux" }:
 
-      with import nixpkgs {inherit system;};
+      with import nixpkgs { inherit system; };
 
       releaseTools.nixBuild {
         name = "nix";
         src = tarball;
 
-        buildInputs = [curl perl bzip2 openssl];
+        buildInputs = [ curl perl bzip2 openssl pkgconfig boehmgc ];
 
         configureFlags = ''
           --disable-init-state
-          --with-aterm=${aterm242fixes} --with-bzip2=${bzip2}
+          --with-bzip2=${bzip2} --with-sqlite=${sqlite}
+          --enable-gc
         '';
       };
 
 
-    /*
-    static =
-      { tarball ? jobs.tarball {}
-      , system ? "i686-linux"
-      }:
-
-      with import nixpkgs {inherit system;};
-
-      releaseTools.binaryTarball {
-        name = "nix-static-tarball";
-        src = tarball;
-
-        buildInputs = [curl perl bzip2];
-
-        configureFlags = ''
-          --disable-init-state
-          --with-aterm=${aterm242fixes} --with-bzip2=${bzip2}
-          --enable-static-nix
-        '';
-      };
-    */
-
-      
     coverage =
-      { tarball ? jobs.tarball {}
-      }:
-
-      with import nixpkgs {};
+      with import nixpkgs { system = "x86_64-linux"; };
 
       releaseTools.coverageAnalysis {
         name = "nix-build";
         src = tarball;
 
-        buildInputs = [
-          curl perl bzip2 openssl
-          # These are for "make check" only:
-          graphviz libxml2 libxslt
-        ];
+        buildInputs =
+          [ curl perl bzip2 openssl
+            # These are for "make check" only:
+            graphviz libxml2 libxslt
+          ];
 
         configureFlags = ''
           --disable-init-state --disable-shared
-          --with-aterm=${aterm242fixes} --with-bzip2=${bzip2}
+          --with-bzip2=${bzip2} --with-sqlite=${sqlite}
         '';
 
-        lcovFilter = ["*/boost/*" "*-tab.*"];
+        lcovFilter = [ "*/boost/*" "*-tab.*" ];
 
         # We call `dot', and even though we just use it to
         # syntax-check generated dot files, it still requires some
@@ -171,17 +143,15 @@ let
 
   makeRPM = 
     system: diskImageFun: prio:
-    { tarball ? jobs.tarball {}
-    }:
 
-    with import nixpkgs {inherit system;};
+    with import nixpkgs { inherit system; };
 
     releaseTools.rpmBuild rec {
       name = "nix-rpm-${diskImage.name}";
-      src = tarball;
+      src = jobs.tarball;
       diskImage = diskImageFun vmTools.diskImages;
       memSize = 1024;
-      meta = { schedulingPriority = toString prio; };
+      meta.schedulingPriority = prio;
     };
 
 
@@ -190,19 +160,17 @@ let
   
   makeDeb =
     system: diskImageFun: prio:
-    { tarball ? jobs.tarball {}
-    }:
 
-    with import nixpkgs {inherit system;};
+    with import nixpkgs { inherit system; };
 
     releaseTools.debBuild {
       name = "nix-deb";
-      src = tarball;
+      src = jobs.tarball;
       diskImage = diskImageFun vmTools.diskImages;
       memSize = 1024;
-      meta = { schedulingPriority = toString prio; };
+      meta.schedulingPriority = prio;
       configureFlags = "--sysconfdir=/etc";
-      debRequires = ["curl"];
+      debRequires = [ "curl" ];
     };
 
 

scripts/GeneratePatches.pm.in

@@ -0,0 +1,334 @@
+#! @perl@ -w -I@libexecdir@/nix
+
+use strict;
+use File::Temp qw(tempdir);
+
+
+# Some patch generations options.
+
+# Max size of NAR archives to generate patches for.
+my $maxNarSize = $ENV{"NIX_MAX_NAR_SIZE"};
+$maxNarSize = 160 * 1024 * 1024 if !defined $maxNarSize;
+
+# If patch is bigger than this fraction of full archive, reject.
+my $maxPatchFraction = $ENV{"NIX_PATCH_FRACTION"};
+$maxPatchFraction = 0.60 if !defined $maxPatchFraction;
+
+my $timeLimit = $ENV{"NIX_BSDIFF_TIME_LIMIT"};
+$timeLimit = 180 if !defined $timeLimit;
+
+my $hashAlgo = "sha256";
+
+
+sub findOutputPaths {
+    my $narFiles = shift;
+
+    my %outPaths;
+    
+    foreach my $p (keys %{$narFiles}) {
+
+        # Ignore derivations.
+        next if ($p =~ /\.drv$/);
+        
+        # Ignore builders (too much ambiguity -- they're all called
+        # `builder.sh').
+        next if ($p =~ /\.sh$/);
+        next if ($p =~ /\.patch$/);
+        
+        # Don't bother including tar files etc.
+        next if ($p =~ /\.tar$/ || $p =~ /\.tar\.(gz|bz2|Z|lzma|xz)$/ || $p =~ /\.zip$/ || $p =~ /\.bin$/ || $p =~ /\.tgz$/ || $p =~ /\.rpm$/ || $p =~ /cvs-export$/ || $p =~ /fetchhg$/);
+
+        $outPaths{$p} = 1;
+    }
+
+    return %outPaths;
+}
+
+
+sub getNameVersion {
+    my $p = shift;
+    $p =~ /\/[0-9a-z]+((?:-[a-zA-Z][^\/-]*)+)([^\/]*)$/;
+    my $name = $1;
+    my $version = $2;
+    return undef unless defined $name && defined $version;
+    $name =~ s/^-//;
+    $version =~ s/^-//;
+    return ($name, $version);
+}
+
+
+# A quick hack to get a measure of the `distance' between two
+# versions: it's just the position of the first character that differs
+# (or 999 if they are the same).
+sub versionDiff {
+    my $s = shift;
+    my $t = shift;
+    my $i;
+    return 999 if $s eq $t;
+    for ($i = 0; $i < length $s; $i++) {
+        return $i if $i >= length $t or
+            substr($s, $i, 1) ne substr($t, $i, 1);
+    }
+    return $i;
+}
+
+
+sub getNarBz2 {
+    my $narPath = shift;
+    my $narFiles = shift;
+    my $storePath = shift;
+    
+    my $narFileList = $$narFiles{$storePath};
+    die "missing path $storePath" unless defined $narFileList;
+
+    my $narFile = @{$narFileList}[0];
+    die unless defined $narFile;
+
+    $narFile->{url} =~ /\/([^\/]+)$/;
+    die unless defined $1;
+    return "$narPath/$1";
+}
+
+
+sub containsPatch {
+    my $patches = shift;
+    my $storePath = shift;
+    my $basePath = shift;
+    my $patchList = $$patches{$storePath};
+    return 0 if !defined $patchList;
+    my $found = 0;
+    foreach my $patch (@{$patchList}) {
+        # !!! baseHash might differ
+        return 1 if $patch->{basePath} eq $basePath;
+    }
+    return 0;
+}
+
+
+sub generatePatches {
+    my ($srcNarFiles, $dstNarFiles, $srcPatches, $dstPatches, $narPath, $patchesPath, $patchesURL, $tmpDir) = @_;
+
+    my %srcOutPaths = findOutputPaths $srcNarFiles;
+    my %dstOutPaths = findOutputPaths $dstNarFiles;
+
+    # For each output path in the destination, see if we need to / can
+    # create a patch.
+
+    print STDERR "creating patches...\n";
+
+    foreach my $p (keys %dstOutPaths) {
+
+        # If exactly the same path already exists in the source, skip it.
+        next if defined $srcOutPaths{$p};
+    
+        print "  $p\n";
+
+        # If not, then we should find the paths in the source that are
+        # `most' likely to be present on a system that wants to
+        # install this path.
+
+        (my $name, my $version) = getNameVersion $p;
+        next unless defined $name && defined $version;
+
+        my @closest = ();
+        my $closestVersion;
+        my $minDist = -1; # actually, larger means closer
+
+        # Find all source paths with the same name.
+
+        foreach my $q (keys %srcOutPaths) {
+            (my $name2, my $version2) = getNameVersion $q;
+            next unless defined $name2 && defined $version2;
+
+            if ($name eq $name2) {
+
+                my $srcSystem = @{$$dstNarFiles{$p}}[0]->{system};
+                my $dstSystem = @{$$srcNarFiles{$q}}[0]->{system};
+                if (defined $srcSystem && defined $dstSystem && $srcSystem ne $dstSystem) {
+                    print "    SKIPPING $q due to different systems ($srcSystem vs. $dstSystem)\n";
+                    next;
+                }
+
+                # If the sizes differ too much, then skip.  This
+                # disambiguates between, e.g., a real component and a
+                # wrapper component (cf. Firefox in Nixpkgs).
+                my $srcSize = @{$$srcNarFiles{$q}}[0]->{size};
+                my $dstSize = @{$$dstNarFiles{$p}}[0]->{size};
+                my $ratio = $srcSize / $dstSize;
+                $ratio = 1 / $ratio if $ratio < 1;
+                # print "  SIZE $srcSize $dstSize $ratio $q\n";
+
+                if ($ratio >= 3) {
+                    print "    SKIPPING $q due to size ratio $ratio ($srcSize vs. $dstSize)\n";
+                    next;
+                }
+
+                # If there are multiple matching names, include the
+                # ones with the closest version numbers.
+                my $dist = versionDiff $version, $version2;
+                if ($dist > $minDist) {
+                    $minDist = $dist;
+                    @closest = ($q);
+                    $closestVersion = $version2;
+                } elsif ($dist == $minDist) {
+                    push @closest, $q;
+                }
+            }
+        }
+
+        if (scalar(@closest) == 0) {
+            print "    NO BASE: $p\n";
+            next;
+        }
+
+        foreach my $closest (@closest) {
+
+            # Generate a patch between $closest and $p.
+            print STDERR "  $p <- $closest\n";
+
+            # If the patch already exists, skip it.
+            if (containsPatch($srcPatches, $p, $closest) ||
+                containsPatch($dstPatches, $p, $closest))
+            {
+                print "    skipping, already exists\n";
+                next;
+            }
+
+            my $srcNarBz2 = getNarBz2 $narPath, $srcNarFiles, $closest;
+            my $dstNarBz2 = getNarBz2 $narPath, $dstNarFiles, $p;
+
+            if (! -f $srcNarBz2) {
+                warn "patch source archive $srcNarBz2 is missing\n";
+                next;
+            }
+
+            system("@bunzip2@ < $srcNarBz2 > $tmpDir/A") == 0
+                or die "cannot unpack $srcNarBz2";
+
+            if ((stat "$tmpDir/A")[7] >= $maxNarSize) {
+                print "    skipping, source is too large\n";
+                next;
+            }
+        
+            system("@bunzip2@ < $dstNarBz2 > $tmpDir/B") == 0
+                or die "cannot unpack $dstNarBz2";
+
+            if ((stat "$tmpDir/B")[7] >= $maxNarSize) {
+                print "    skipping, destination is too large\n";
+                next;
+            }
+        
+            my $time1 = time();
+            my $res = system("ulimit -t $timeLimit; @libexecdir@/bsdiff $tmpDir/A $tmpDir/B $tmpDir/DIFF");
+            my $time2 = time();
+            if ($res) {
+                warn "binary diff computation aborted after ", $time2 - $time1, " seconds\n";
+                next;
+            }
+
+            my $baseHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/A` or die;
+            chomp $baseHash;
+
+            my $narHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/B` or die;
+            chomp $narHash;
+
+            my $narDiffHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/DIFF` or die;
+            chomp $narDiffHash;
+
+            my $narDiffSize = (stat "$tmpDir/DIFF")[7];
+            my $dstNarBz2Size = (stat $dstNarBz2)[7];
+
+            print "    size $narDiffSize; full size $dstNarBz2Size; ", $time2 - $time1, " seconds\n";
+        
+            if ($narDiffSize >= $dstNarBz2Size) {
+                print "    rejecting; patch bigger than full archive\n";
+                next;
+            }
+    
+            if ($narDiffSize / $dstNarBz2Size >= $maxPatchFraction) {
+                print "    rejecting; patch too large relative to full archive\n";
+                next;
+            }
+    
+            my $finalName = "$narDiffHash.nar-bsdiff";
+
+            if (-e "$patchesPath/$finalName") {
+                print "    not copying, already exists\n";
+            }
+
+            else {
+                system("cp '$tmpDir/DIFF' '$patchesPath/$finalName.tmp'") == 0
+                    or die "cannot copy diff";
+                rename("$patchesPath/$finalName.tmp", "$patchesPath/$finalName")
+                    or die "cannot rename $patchesPath/$finalName.tmp";
+            }
+        
+            # Add the patch to the manifest.
+            addPatch $dstPatches, $p,
+                { url => "$patchesURL/$finalName", hash => "$hashAlgo:$narDiffHash"
+                , size => $narDiffSize, basePath => $closest, baseHash => "$hashAlgo:$baseHash"
+                , narHash => "$hashAlgo:$narHash", patchType => "nar-bsdiff"
+                };
+        }
+    }
+}
+
+
+# Propagate useful patches from $srcPatches to $dstPatches.  A patch
+# is useful if it produces either paths in the $dstNarFiles or paths
+# that can be used as the base for other useful patches.
+sub propagatePatches {
+    my ($srcPatches, $dstNarFiles, $dstPatches) = @_;
+
+    print STDERR "propagating patches...\n";
+
+    my $changed;
+    do {
+        # !!! we repeat this to reach the transitive closure; inefficient
+        $changed = 0;
+
+        print STDERR "loop\n";
+
+        my %dstBasePaths;
+        foreach my $q (keys %{$dstPatches}) {
+            foreach my $patch (@{$$dstPatches{$q}}) {
+                $dstBasePaths{$patch->{basePath}} = 1;
+            }
+        }
+
+        foreach my $p (keys %{$srcPatches}) {
+            my $patchList = $$srcPatches{$p};
+
+            my $include = 0;
+
+            # Is path $p included in the destination?  If so, include
+            # patches that produce it.
+            $include = 1 if defined $$dstNarFiles{$p};
+
+            # Is path $p a path that serves as a base for paths in the
+            # destination?  If so, include patches that produce it.
+            # !!! check baseHash
+            $include = 1 if defined $dstBasePaths{$p};
+
+            if ($include) {
+                foreach my $patch (@{$patchList}) {
+                    $changed = 1 if addPatch $dstPatches, $p, $patch;
+                }
+            }
+        
+        }
+    
+    } while $changed;
+}
+
+
+# Add all new patches in $srcPatches to $dstPatches.
+sub copyPatches {
+    my ($srcPatches, $dstPatches) = @_;
+    foreach my $p (keys %{$srcPatches}) {
+        addPatch $dstPatches, $p, $_ foreach @{$$srcPatches{$p}};
+    }
+}
+
+
+return 1;

scripts/Makefile.am

@@ -1,23 +1,23 @@
 bin_SCRIPTS = nix-collect-garbage \
   nix-pull nix-push nix-prefetch-url \
   nix-install-package nix-channel nix-build \
-  nix-copy-closure 
+  nix-copy-closure nix-generate-patches
 
-noinst_SCRIPTS = nix-profile.sh generate-patches.pl \
+noinst_SCRIPTS = nix-profile.sh GeneratePatches.pm \
   find-runtime-roots.pl build-remote.pl nix-reduce-build \
   copy-from-other-stores.pl nix-http-export.cgi
 
-nix-pull nix-push: readmanifest.pm readconfig.pm download-using-manifests.pl
+nix-pull nix-push: NixManifest.pm NixConfig.pm download-using-manifests.pl
 
-install-exec-local: readmanifest.pm download-using-manifests.pl copy-from-other-stores.pl find-runtime-roots.pl
+install-exec-local: NixManifest.pm GeneratePatches.pm download-using-manifests.pl copy-from-other-stores.pl find-runtime-roots.pl
 	$(INSTALL) -d $(DESTDIR)$(sysconfdir)/profile.d
 	$(INSTALL_PROGRAM) nix-profile.sh $(DESTDIR)$(sysconfdir)/profile.d/nix.sh
 	$(INSTALL) -d $(DESTDIR)$(libexecdir)/nix
-	$(INSTALL_DATA) readmanifest.pm $(DESTDIR)$(libexecdir)/nix 
-	$(INSTALL_DATA) readconfig.pm $(DESTDIR)$(libexecdir)/nix 
-	$(INSTALL_DATA) ssh.pm $(DESTDIR)$(libexecdir)/nix 
+	$(INSTALL_DATA) NixManifest.pm $(DESTDIR)$(libexecdir)/nix 
+	$(INSTALL_DATA) NixConfig.pm $(DESTDIR)$(libexecdir)/nix 
+	$(INSTALL_DATA) SSH.pm $(DESTDIR)$(libexecdir)/nix 
+	$(INSTALL_DATA) GeneratePatches.pm $(DESTDIR)$(libexecdir)/nix 
 	$(INSTALL_PROGRAM) find-runtime-roots.pl $(DESTDIR)$(libexecdir)/nix 
-	$(INSTALL_PROGRAM) generate-patches.pl $(DESTDIR)$(libexecdir)/nix 
 	$(INSTALL_PROGRAM) build-remote.pl $(DESTDIR)$(libexecdir)/nix 
 	$(INSTALL) -d $(DESTDIR)$(libexecdir)/nix/substituters
 	$(INSTALL_PROGRAM) download-using-manifests.pl $(DESTDIR)$(libexecdir)/nix/substituters
@@ -30,15 +30,16 @@ EXTRA_DIST = nix-collect-garbage.in \
   nix-pull.in nix-push.in nix-profile.sh.in \
   nix-prefetch-url.in nix-install-package.in \
   nix-channel.in \
-  readmanifest.pm.in \
-  readconfig.pm.in \
-  ssh.pm \
+  NixManifest.pm.in \
+  NixConfig.pm.in \
+  SSH.pm \
+  GeneratePatches.pm.in \
   nix-build.in \
   download-using-manifests.pl.in \
   copy-from-other-stores.pl.in \
-  generate-patches.pl.in \
   nix-copy-closure.in \
   find-runtime-roots.pl.in \
   build-remote.pl.in \
   nix-reduce-build.in \
-  nix-http-export.cgi.in 
+  nix-http-export.cgi.in \
+  nix-generate-patches.in

scripts/NixManifest.pm.in

@@ -33,18 +33,8 @@ sub readManifest {
 
     my $manifestVersion = 2;
 
-    my $storePath;
-    my $url;
-    my $hash;
-    my $size;
-    my $basePath;
-    my $baseHash;
-    my $patchType;
-    my $narHash;
-    my $references;
-    my $deriver;
-    my $hashAlgo;
-    my $copyFrom;
+    my ($storePath, $url, $hash, $size, $basePath, $baseHash, $patchType);
+    my ($narHash, $narSize, $references, $deriver, $hashAlgo, $copyFrom, $system);
 
     while (<MANIFEST>) {
         chomp;
@@ -62,9 +52,11 @@ sub readManifest {
                 undef $hash;
                 undef $size;
                 undef $narHash;
+                undef $narSize;
                 undef $basePath;
                 undef $baseHash;
                 undef $patchType;
+                undef $system;
                 $references = "";
                 $deriver = "";
                 $hashAlgo = "md5";
@@ -89,8 +81,10 @@ sub readManifest {
                     if (!$found) {
                         push @{$narFileList},
                             { url => $url, hash => $hash, size => $size
-                            , narHash => $narHash, references => $references
+                            , narHash => $narHash, narSize => $narSize
+                            , references => $references
                             , deriver => $deriver, hashAlgo => $hashAlgo
+                            , system => $system
                             };
                     }
                 
@@ -100,8 +94,8 @@ sub readManifest {
                     addPatch $patches, $storePath,
                         { url => $url, hash => $hash, size => $size
                         , basePath => $basePath, baseHash => $baseHash
-                        , narHash => $narHash, patchType => $patchType
-                        , hashAlgo => $hashAlgo
+                        , narHash => $narHash, narSize => $narSize
+                        , patchType => $patchType, hashAlgo => $hashAlgo
                         };
                 }
 
@@ -132,9 +126,11 @@ sub readManifest {
             elsif (/^\s*BaseHash:\s*(\S+)\s*$/) { $baseHash = $1; }
             elsif (/^\s*Type:\s*(\S+)\s*$/) { $patchType = $1; }
             elsif (/^\s*NarHash:\s*(\S+)\s*$/) { $narHash = $1; }
+            elsif (/^\s*NarSize:\s*(\d+)\s*$/) { $narSize = $1; }
             elsif (/^\s*References:\s*(.*)\s*$/) { $references = $1; }
             elsif (/^\s*Deriver:\s*(\S+)\s*$/) { $deriver = $1; }
             elsif (/^\s*ManifestVersion:\s*(\d+)\s*$/) { $manifestVersion = $1; }
+            elsif (/^\s*System:\s*(\S+)\s*$/) { $system = $1; }
 
             # Compatibility;
             elsif (/^\s*NarURL:\s*(\S+)\s*$/) { $url = $1; }
@@ -150,7 +146,7 @@ sub readManifest {
 
 
 sub writeManifest {
-    my ($manifest, $narFiles, $patches) = @_;
+    my ($manifest, $narFiles, $patches, $noCompress) = @_;
 
     open MANIFEST, ">$manifest.tmp"; # !!! check exclusive
 
@@ -165,12 +161,14 @@ sub writeManifest {
             print MANIFEST "  StorePath: $storePath\n";
             print MANIFEST "  NarURL: $narFile->{url}\n";
             print MANIFEST "  Hash: $narFile->{hash}\n" if defined $narFile->{hash};
-            print MANIFEST "  NarHash: $narFile->{narHash}\n";
             print MANIFEST "  Size: $narFile->{size}\n" if defined $narFile->{size};
+            print MANIFEST "  NarHash: $narFile->{narHash}\n";
+            print MANIFEST "  NarSize: $narFile->{narSize}\n" if $narFile->{narSize};
             print MANIFEST "  References: $narFile->{references}\n"
                 if defined $narFile->{references} && $narFile->{references} ne "";
             print MANIFEST "  Deriver: $narFile->{deriver}\n"
                 if defined $narFile->{deriver} && $narFile->{deriver} ne "";
+            print MANIFEST "  System: $narFile->{system}\n" if defined $narFile->{system};
             print MANIFEST "}\n";
         }
     }
@@ -182,8 +180,9 @@ sub writeManifest {
             print MANIFEST "  StorePath: $storePath\n";
             print MANIFEST "  NarURL: $patch->{url}\n";
             print MANIFEST "  Hash: $patch->{hash}\n";
-            print MANIFEST "  NarHash: $patch->{narHash}\n";
             print MANIFEST "  Size: $patch->{size}\n";
+            print MANIFEST "  NarHash: $patch->{narHash}\n";
+            print MANIFEST "  NarSize: $patch->{narSize}\n" if $patch->{narSize};
             print MANIFEST "  BasePath: $patch->{basePath}\n";
             print MANIFEST "  BaseHash: $patch->{baseHash}\n";
             print MANIFEST "  Type: $patch->{patchType}\n";
@@ -199,11 +198,13 @@ sub writeManifest {
 
 
     # Create a bzipped manifest.
-    system("@bzip2@ < $manifest > $manifest.bz2.tmp") == 0
-        or die "cannot compress manifest";
+    unless (defined $noCompress) {
+	system("@bzip2@ < $manifest > $manifest.bz2.tmp") == 0
+	    or die "cannot compress manifest";
 
-    rename("$manifest.bz2.tmp", "$manifest.bz2")
-        or die "cannot rename $manifest.bz2.tmp: $!";
+	rename("$manifest.bz2.tmp", "$manifest.bz2")
+	    or die "cannot rename $manifest.bz2.tmp: $!";
+    }
 }
 
 

scripts/SSH.pm

@@ -3,6 +3,8 @@ use File::Temp qw(tempdir);
 
 our @sshOpts = split ' ', ($ENV{"NIX_SSHOPTS"} or "");
 
+push @sshOpts, "-x";
+
 my $sshStarted = 0;
 my $sshHost;
 
@@ -24,14 +26,17 @@ sub openSSHConnection {
     # child continues to run if we are killed.  So instead make SSH
     # print "started" when it has established the connection, and wait
     # until we see that.
-    open SSH, "ssh $sshHost @sshOpts -M -N -o LocalCommand='echo started' -o PermitLocalCommand=yes |" or die;
-    while (<SSH>) {
+    open SSHPIPE, "ssh $sshHost @sshOpts -M -N -o LocalCommand='echo started' -o PermitLocalCommand=yes |" or die;
+
+    while (<SSHPIPE>) {
         chomp;
-        last if /started/;
+        if ($_ eq "started") {
+            $sshStarted = 1;
+            return 1;
+        }
     }
-    
-    $sshStarted = 1;
-    return 1;
+
+    return 0;
 }
 
 # Tell the master SSH client to exit.

scripts/build-remote.pl.in

@@ -3,7 +3,8 @@
 use Fcntl ':flock';
 use English '-no_match_vars';
 use IO::Handle;
-use ssh qw/sshOpts openSSHConnection/;
+use SSH qw/sshOpts openSSHConnection/;
+no warnings('once');
 
 
 # General operation:
@@ -31,57 +32,22 @@ $ENV{"DISPLAY"} = "";
 $ENV{"SSH_ASKPASS"} = "";
 
 
-my $loadIncreased = 0;
-
-my ($amWilling, $localSystem, $neededSystem, $drvPath, $maxSilentTime) = @ARGV;
-$maxSilentTime = 0 unless defined $maxSilentTime;
-
 sub sendReply {
     my $reply = shift;
     print STDERR "# $reply\n";
 }
 
-sub decline {
-    sendReply "decline";
-    exit 0;
-}
+sub all { $_ || return 0 for @_; 1 }
+
+
+# Initialisation.
+my $loadIncreased = 0;
+
+my ($localSystem, $maxSilentTime, $printBuildTrace) = @ARGV;
+$maxSilentTime = 0 unless defined $maxSilentTime;
 
 my $currentLoad = $ENV{"NIX_CURRENT_LOAD"};
-decline unless defined $currentLoad;
-mkdir $currentLoad, 0777 or die unless -d $currentLoad;
-
 my $conf = $ENV{"NIX_REMOTE_SYSTEMS"};
-decline if !defined $conf || ! -e $conf;
-
-my $canBuildLocally = $amWilling && ($localSystem eq $neededSystem);
-
-
-# Read the list of machines.
-my @machines;
-open CONF, "< $conf" or die;
-
-while (<CONF>) {
-    chomp;
-    s/\#.*$//g;
-    next if /^\s*$/;
-    /^\s*(\S+)\s+(\S+)\s+(\S+)\s+(\d+)(\s+([0-9\.]+))?\s*$/ or die;
-    push @machines,
-        { hostName => $1
-        , systemTypes => [split(/,/, $2)]
-        , sshKeys => $3
-        , maxJobs => $4
-        , speedFactor => 1.0 * ($6 || 1)
-        , enabled => 1
-        };
-}
-
-close CONF;
-
-
-# Acquire the exclusive lock on $currentLoad/main-lock.
-my $mainLock = "$currentLoad/main-lock";
-open MAINLOCK, ">>$mainLock" or die;
-flock(MAINLOCK, LOCK_EX) or die;
 
 
 sub openSlotLock {
@@ -91,150 +57,213 @@ sub openSlotLock {
     open $slotLock, ">>$slotLockFn" or die;
     return $slotLock;
 }
+
+
+# Read the list of machines.
+my @machines;
+if (defined $conf && -e $conf) {
+    open CONF, "< $conf" or die;
+    while (<CONF>) {
+        chomp;
+        s/\#.*$//g;
+        next if /^\s*$/;
+        my @tokens = split /\s/, $_;
+        push @machines,
+            { hostName => $tokens[0]
+            , systemTypes => [ split(/,/, $tokens[1]) ]
+            , sshKeys => $tokens[2]
+            , maxJobs => int($tokens[3])
+            , speedFactor => 1.0 * (defined $tokens[4] ? int($tokens[4]) : 1)
+            , features => [ split(/,/, $tokens[5] || "") ]
+            , enabled => 1
+            };
+    }
+    close CONF;
+}
+
+
+
+# Wait for the calling process to ask us whether we can build some derivation.
+my ($drvPath, $hostName, $slotLock);
+
+REQ: while (1) {
+    $_ = <STDIN> || exit 0;
+    my ($amWilling, $neededSystem);
+    ($amWilling, $neededSystem, $drvPath, $requiredFeatures) = split;
+    my @requiredFeatures = split /,/, $requiredFeatures;
+
+    my $canBuildLocally = $amWilling && ($localSystem eq $neededSystem);
+
+    if (!defined $currentLoad) {
+        sendReply "decline";
+        next;
+    }
     
-
-my $hostName;
-my $slotLock;
-
-while (1) {
+    # Acquire the exclusive lock on $currentLoad/main-lock.
+    mkdir $currentLoad, 0777 or die unless -d $currentLoad;
+    my $mainLock = "$currentLoad/main-lock";
+    open MAINLOCK, ">>$mainLock" or die;
+    flock(MAINLOCK, LOCK_EX) or die;
     
-    # Find all machine that can execute this build, i.e., that support
-    # builds for the given platform and are not at their job limit.
-    my $rightType = 0;
-    my @available = ();
-    LOOP: foreach my $cur (@machines) {
-        if ($cur->{enabled} && grep { $neededSystem eq $_ } @{$cur->{systemTypes}}) {
-            $rightType = 1;
+    
+    while (1) {
+        # Find all machine that can execute this build, i.e., that
+        # support builds for the given platform and features, and are
+        # not at their job limit.
+        my $rightType = 0;
+        my @available = ();
+        LOOP: foreach my $cur (@machines) {
+            if ($cur->{enabled}
+                && (grep { $neededSystem eq $_ } @{$cur->{systemTypes}})
+                && all(map { my $f = $_; 0 != grep { $f eq $_ } @{$cur->{features}} } @requiredFeatures))
+            {
+                $rightType = 1;
 
-            # We have a machine of the right type.  Determine the load on
-            # the machine.
-            my $slot = 0;
-            my $load = 0;
-            my $free;
-            while ($slot < $cur->{maxJobs}) {
-                my $slotLock = openSlotLock($cur, $slot);
-                if (flock($slotLock, LOCK_EX | LOCK_NB)) {
-                    $free = $slot unless defined $free;
-                    flock($slotLock, LOCK_UN) or die;
-                } else {
-                    $load++;
+                # We have a machine of the right type.  Determine the load on
+                # the machine.
+                my $slot = 0;
+                my $load = 0;
+                my $free;
+                while ($slot < $cur->{maxJobs}) {
+                    my $slotLock = openSlotLock($cur, $slot);
+                    if (flock($slotLock, LOCK_EX | LOCK_NB)) {
+                        $free = $slot unless defined $free;
+                        flock($slotLock, LOCK_UN) or die;
+                    } else {
+                        $load++;
+                    }
+                    close $slotLock;
+                    $slot++;
                 }
-                close $slotLock;
-                $slot++;
+                
+                push @available, { machine => $cur, load => $load, free => $free }
+                if $load < $cur->{maxJobs};
             }
-
-            push @available, { machine => $cur, load => $load, free => $free }
-            if $load < $cur->{maxJobs};
         }
-    }
 
-    if (defined $ENV{NIX_DEBUG_HOOK}) {
-        print STDERR "load on " . $_->{machine}->{hostName} . " = " . $_->{load} . "\n"
-            foreach @available;
-    }
-
-
-    # Didn't find any available machine?  Then decline or postpone.
-    if (scalar @available == 0) {
-        # Postpone if we have a machine of the right type, except if the
-        # local system can and wants to do the build.
-        if ($rightType && !$canBuildLocally) {
-            sendReply "postpone";
-            exit 0;
-        } else {
-            decline;
+        if (defined $ENV{NIX_DEBUG_HOOK}) {
+            print STDERR "load on " . $_->{machine}->{hostName} . " = " . $_->{load} . "\n"
+                foreach @available;
         }
-    }
 
 
-    # Prioritise the available machines as follows:
-    # - First by load divided by speed factor, rounded to the nearest
-    #   integer.  This causes fast machines to be preferred over slow
-    #   machines with similar loads.
-    # - Then by speed factor.
-    # - Finally by load.
-    sub lf { my $x = shift; return int($x->{load} / $x->{machine}->{speedFactor} + 0.4999); }
-    @available = sort
-        { lf($a) <=> lf($b)
-              || $b->{machine}->{speedFactor} <=> $a->{machine}->{speedFactor}
-              || $a->{load} <=> $b->{load}
-        } @available;
+        # Didn't find any available machine?  Then decline or postpone.
+        if (scalar @available == 0) {
+            # Postpone if we have a machine of the right type, except
+            # if the local system can and wants to do the build.
+            if ($rightType && !$canBuildLocally) {
+                sendReply "postpone";
+            } else {
+                sendReply "decline";                
+            }
+            close MAINLOCK;
+            next REQ;
+        }
 
 
-    # Select the best available machine and lock a free slot.
-    my $selected = $available[0]; 
-    my $machine = $selected->{machine};
-
-    $slotLock = openSlotLock($machine, $selected->{free});
-    flock($slotLock, LOCK_EX | LOCK_NB) or die;
-    utime undef, undef, $slotLock;
-
-    close MAINLOCK;
+        # Prioritise the available machines as follows:
+        # - First by load divided by speed factor, rounded to the nearest
+        #   integer.  This causes fast machines to be preferred over slow
+        #   machines with similar loads.
+        # - Then by speed factor.
+        # - Finally by load.
+        sub lf { my $x = shift; return int($x->{load} / $x->{machine}->{speedFactor} + 0.4999); }
+        @available = sort
+            { lf($a) <=> lf($b)
+                  || $b->{machine}->{speedFactor} <=> $a->{machine}->{speedFactor}
+                  || $a->{load} <=> $b->{load}
+            } @available;
 
 
-    # Connect to the selected machine.
-    @sshOpts = ("-i", $machine->{sshKeys}, "-x");
-    $hostName = $machine->{hostName};
-    last if openSSHConnection $hostName;
+        # Select the best available machine and lock a free slot.
+        my $selected = $available[0]; 
+        my $machine = $selected->{machine};
+        
+        $slotLock = openSlotLock($machine, $selected->{free});
+        flock($slotLock, LOCK_EX | LOCK_NB) or die;
+        utime undef, undef, $slotLock;
+
+        close MAINLOCK;
+
+
+        # Connect to the selected machine.
+        @sshOpts = ("-i", $machine->{sshKeys}, "-x");
+        $hostName = $machine->{hostName};
+        last REQ if openSSHConnection $hostName;
     
-    warn "unable to open SSH connection to $hostName, trying other available machines...\n";
-    $machine->{enabled} = 0;
+        warn "unable to open SSH connection to $hostName, trying other available machines...\n";
+        $machine->{enabled} = 0;
+    }
 }
 
 
 # Tell Nix we've accepted the build.
 sendReply "accept";
-my $x = <STDIN>;
-chomp $x;
-
-if ($x ne "okay") {
-    exit 0;
-}
+my @inputs = split /\s/, readline(STDIN);
+my @outputs = split /\s/, readline(STDIN);
 
 
-# Do the actual build.
 print STDERR "building `$drvPath' on `$hostName'\n";
+print STDERR "@ build-remote $drvPath $hostName\n" if $printBuildTrace;
 
-my $inputs = `cat inputs`; die if ($? != 0);
-$inputs =~ s/\n/ /g;
-
-my $outputs = `cat outputs`; die if ($? != 0);
-$outputs =~ s/\n/ /g;
-
-print "copying inputs...\n";
 
 my $maybeSign = "";
 $maybeSign = "--sign" if -e "/nix/etc/nix/signing-key.sec";
 
-system("NIX_SSHOPTS=\"@sshOpts\" @bindir@/nix-copy-closure $hostName $maybeSign $drvPath $inputs") == 0
+
+# Register the derivation as a temporary GC root.  Note that $PPID is
+# the PID of the remote SSH process, which, due to the use of a
+# persistant SSH connection, should be the same across all remote
+# command invocations for this session.
+my $rootsDir = "@localstatedir@/nix/gcroots/tmp";
+system("ssh $hostName @sshOpts 'mkdir -m 1777 -p $rootsDir; ln -sfn $drvPath $rootsDir/\$PPID.drv'");
+
+sub removeRoots {
+    system("ssh $hostName @sshOpts 'rm -f $rootsDir/\$PPID.drv $rootsDir/\$PPID.out'");
+}
+
+
+# Copy the derivation and its dependencies to the build machine.
+system("NIX_SSHOPTS=\"@sshOpts\" @bindir@/nix-copy-closure $hostName $maybeSign $drvPath @inputs") == 0
     or die "cannot copy inputs to $hostName: $?";
 
-print "building...\n";
 
-my $buildFlags = "--max-silent-time $maxSilentTime";
+# Perform the build.
+my $buildFlags = "--max-silent-time $maxSilentTime --fallback --add-root $rootsDir/\$PPID.out --option verbosity 0";
 
-# `-tt' forces allocation of a pseudo-terminal.  This is required to
-# make the remote nix-store process receive a signal when the
-# connection dies.  Without it, the remote process might continue to
-# run indefinitely (that is, until it next tries to write to
-# stdout/stderr).
-if (system("ssh $hostName @sshOpts -tt 'nix-store --realise $buildFlags $drvPath > /dev/null'") != 0) {
-    # If we couldn't run ssh or there was an ssh problem (indicated by
-    # exit code 255), then we return exit code 1; otherwise we assume
-    # that the builder failed, which we indicate to Nix using exit
-    # code 100.  It's important to distinguish between the two because
-    # the first is a transient failure and the latter is permanent.
-    my $res = $? == -1 || ($? >> 8) == 255 ? 1 : 100;
-    print STDERR "build of `$drvPath' on `$hostName' failed with exit code $?\n";
+# We let the remote side kill its process group when the connection is
+# closed unexpectedly.  This is necessary to ensure that no processes
+# are left running on the remote system if the local Nix process is
+# killed.  (SSH itself doesn't kill child processes if the connection
+# is interrupted unless the `-tt' flag is used to force a pseudo-tty,
+# in which case every child receives SIGHUP; however, `-tt' doesn't
+# work on some platforms when connection sharing is used.)
+pipe STDIN, DUMMY; # make sure we have a readable STDIN
+if (system("ssh $hostName @sshOpts '(read; kill -INT -\$\$) <&0 & nix-store -r $drvPath $buildFlags > /dev/null' 2>&4") != 0) {
+    # Note that if we get exit code 100 from `nix-store -r', it
+    # denotes a permanent build failure (as opposed to an SSH problem
+    # or a temporary Nix problem).  We propagate this to the caller to
+    # allow it to distinguish between transient and permanent
+    # failures.
+    my $res = $? >> 8;
+    print STDERR "build of `$drvPath' on `$hostName' failed with exit code $res\n";
+    removeRoots;
     exit $res;
 }
 
-print "build of `$drvPath' on `$hostName' succeeded\n";
+#print "build of `$drvPath' on `$hostName' succeeded\n";
 
-foreach my $output (split '\n', $outputs) {
+
+# Copy the output from the build machine.
+foreach my $output (@outputs) {
     my $maybeSignRemote = "";
     $maybeSignRemote = "--sign" if $UID != 0;
     
-    system("ssh $hostName @sshOpts 'nix-store --export $maybeSignRemote $output' | @bindir@/nix-store --import > /dev/null") == 0
+    system("ssh $hostName @sshOpts 'nix-store --export $maybeSignRemote $output'" .
+           "| NIX_HELD_LOCKS=$output @bindir@/nix-store --import > /dev/null") == 0
 	or die "cannot copy $output from $hostName: $?";
 }
+
+
+# Get rid of the temporary GC roots.
+removeRoots;

scripts/copy-from-other-stores.pl.in

@@ -17,25 +17,19 @@ foreach my $dir (@remoteStoresAll) {
 }
 
 
+$ENV{"NIX_REMOTE"} = "";
+
+
 sub findStorePath {
     my $storePath = shift;
-    
-    my $storePathName = basename $storePath;
-    
     foreach my $store (@remoteStores) {
-        # Determine whether $storePath exists by looking for the
-        # existence of the info file, and if so, get store path info
-        # from that file.  This rather breaks abstraction: we should
-        # be using `nix-store' for that.  But right now there is no
-        # good way to tell nix-store to access a store mounted under a
-        # different location (there's $NIX_STORE, but that only works
-        # if the remote store is mounted under its "real" location).
-        my $infoFile = "$store/var/nix/db/info/$storePathName";
-        my $storePath2 = "$store/store/$storePathName";
-        if (-f $infoFile && -e $storePath2) {
-            return ($infoFile, $storePath2);
-        }
+        my $sourcePath = "$store/store/" . basename $storePath;
+        next unless -e $sourcePath || -l $sourcePath;
+        $ENV{"NIX_DB_DIR"} = "$store/var/nix/db";
+        return ($store, $sourcePath) if
+            system("@bindir@/nix-store --check-validity $storePath") == 0;
     }
+    return undef;
 }
 
 
@@ -46,37 +40,38 @@ if ($ARGV[0] eq "--query") {
 
         if ($cmd eq "have") {
             my $storePath = <STDIN>; chomp $storePath;
-            (my $infoFile) = findStorePath $storePath;
-            print STDOUT ($infoFile ? "1\n" : "0\n");
+            print STDOUT (defined findStorePath($storePath) ? "1\n" : "0\n");
         }
 
         elsif ($cmd eq "info") {
             my $storePath = <STDIN>; chomp $storePath;
-            (my $infoFile) = findStorePath $storePath;
-            if (!$infoFile) {
+            my ($store, $sourcePath) = findStorePath($storePath);
+            if (!defined $store) {
                 print "0\n";
                 next; # not an error
             }
             print "1\n";
 
-            my $deriver = "";
-            my @references = ();
+            $ENV{"NIX_DB_DIR"} = "$store/var/nix/db";
+            
+            my $deriver = `@bindir@/nix-store --query --deriver $storePath`;
+            die "cannot query deriver of `$storePath'" if $? != 0;
+            chomp $deriver;
+            $deriver = "" if $deriver eq "unknown-deriver";
 
-            open INFO, "<$infoFile" or die "cannot read info file $infoFile\n";
-            while (<INFO>) {
-                chomp;
-                /^([\w-]+): (.*)$/ or die "bad info file";
-                my $key = $1;
-                my $value = $2;
-                if ($key eq "Deriver") { $deriver = $value; }
-                elsif ($key eq "References") { @references = split ' ', $value; }
-            }
-            close INFO;
+            my @references = split "\n",
+                `@bindir@/nix-store --query --references $storePath`;
+            die "cannot query references of `$storePath'" if $? != 0;
+
+            my $narSize = `@bindir@/nix-store --query --size $storePath`;
+            die "cannot query size of `$storePath'" if $? != 0;
+            chomp $narSize;
 
             print "$deriver\n";
             print scalar @references, "\n";
             print "$_\n" foreach @references;
-            print "0\n"; # !!! showing size not supported (yet)
+            print "$narSize\n";
+            print "$narSize\n";
         }
 
         else { die "unknown command `$cmd'"; }
@@ -87,8 +82,8 @@ if ($ARGV[0] eq "--query") {
 elsif ($ARGV[0] eq "--substitute") {
     die unless scalar @ARGV == 2;
     my $storePath = $ARGV[1];
-    (my $infoFile, my $sourcePath) = findStorePath $storePath;
-    die unless $infoFile;
+    my ($store, $sourcePath) = findStorePath $storePath;
+    die unless $store;
     print "\n*** Copying `$storePath' from `$sourcePath'\n\n";
     system("$binDir/nix-store --dump $sourcePath | $binDir/nix-store --restore $storePath") == 0
         or die "cannot copy `$sourcePath' to `$storePath'";

scripts/download-using-manifests.pl.in

@@ -1,7 +1,7 @@
 #! @perl@ -w -I@libexecdir@/nix
 
 use strict;
-use readmanifest;
+use NixManifest;
 use POSIX qw(strftime);
 use File::Temp qw(tempdir);
 
@@ -12,6 +12,10 @@ STDOUT->autoflush(1);
 my $manifestDir = ($ENV{"NIX_MANIFESTS_DIR"} or "@localstatedir@/nix/manifests");
 my $logFile = "@localstatedir@/log/nix/downloads";
 
+# For queries, skip expensive calls to nix-hash etc.  We're just
+# estimating the expected download size.
+my $fast = 1;
+
 
 # Load all manifests.
 my %narFiles;
@@ -31,6 +35,151 @@ for my $manifest (glob "$manifestDir/*.nixmanifest") {
 }
 
 
+sub isValidPath {
+    my $p = shift;
+    if ($fast) {
+        return -e $p;
+    } else {
+        return system("$binDir/nix-store --check-validity '$p' 2> /dev/null") == 0;
+    }
+}
+
+
+sub parseHash {
+    my $hash = shift;
+    if ($hash =~ /^(.+):(.+)$/) {
+        return ($1, $2);
+    } else {
+        return ("md5", $hash);
+    }
+}
+
+
+# Compute the most efficient sequence of downloads to produce the
+# given path.
+sub computeSmallestDownload {
+    my $targetPath = shift;
+    
+    # Build a graph of all store paths that might contribute to the
+    # construction of $targetPath, and the special node "start".  The
+    # edges are either patch operations, or downloads of full NAR
+    # files.  The latter edges only occur between "start" and a store
+    # path.
+    my %graph;
+
+    $graph{"start"} = {d => 0, pred => undef, edges => []};
+
+    my @queue = ();
+    my $queueFront = 0;
+    my %done;
+
+    sub addNode {
+        my $graph = shift;
+        my $u = shift;
+        $$graph{$u} = {d => 999999999999, pred => undef, edges => []}
+            unless defined $$graph{$u};
+    }
+
+    sub addEdge {
+        my $graph = shift;
+        my $u = shift;
+        my $v = shift;
+        my $w = shift;
+        my $type = shift;
+        my $info = shift;
+        addNode $graph, $u;
+        push @{$$graph{$u}->{edges}},
+            {weight => $w, start => $u, end => $v, type => $type, info => $info};
+        my $n = scalar @{$$graph{$u}->{edges}};
+    }
+
+    push @queue, $targetPath;
+
+    while ($queueFront < scalar @queue) {
+        my $u = $queue[$queueFront++];
+        next if defined $done{$u};
+        $done{$u} = 1;
+
+        addNode \%graph, $u;
+
+        # If the path already exists, it has distance 0 from the
+        # "start" node.
+        if (isValidPath($u)) {
+            addEdge \%graph, "start", $u, 0, "present", undef;
+        }
+
+        else {
+
+            # Add patch edges.
+            my $patchList = $patches{$u};
+            foreach my $patch (@{$patchList}) {
+                if (isValidPath($patch->{basePath})) {
+                    # !!! this should be cached
+                    my ($baseHashAlgo, $baseHash) = parseHash $patch->{baseHash};
+                    my $format = "--base32";
+                    $format = "" if $baseHashAlgo eq "md5";
+                    my $hash = $fast && $baseHashAlgo eq "sha256"
+                        ? `$binDir/nix-store -q --hash "$patch->{basePath}"`
+                        : `$binDir/nix-hash --type '$baseHashAlgo' $format "$patch->{basePath}"`;
+                    chomp $hash;
+                    $hash =~ s/.*://;
+                    next if $hash ne $baseHash;
+                }
+                push @queue, $patch->{basePath};
+                addEdge \%graph, $patch->{basePath}, $u, $patch->{size}, "patch", $patch;
+            }
+
+            # Add NAR file edges to the start node.
+            my $narFileList = $narFiles{$u};
+            foreach my $narFile (@{$narFileList}) {
+                # !!! how to handle files whose size is not known in advance?
+                # For now, assume some arbitrary size (1 MB).
+                addEdge \%graph, "start", $u, ($narFile->{size} || 1000000), "narfile", $narFile;
+            }
+        }
+    }
+
+
+    # Run Dijkstra's shortest path algorithm to determine the shortest
+    # sequence of download and/or patch actions that will produce
+    # $targetPath.
+
+    my @todo = keys %graph;
+
+    while (scalar @todo > 0) {
+
+        # Remove the closest element from the todo list.
+        # !!! inefficient, use a priority queue
+        @todo = sort { -($graph{$a}->{d} <=> $graph{$b}->{d}) } @todo;
+        my $u = pop @todo;
+
+        my $u_ = $graph{$u};
+
+        foreach my $edge (@{$u_->{edges}}) {
+            my $v_ = $graph{$edge->{end}};
+            if ($v_->{d} > $u_->{d} + $edge->{weight}) {
+                $v_->{d} = $u_->{d} + $edge->{weight};
+                # Store the edge; to edge->start is actually the
+                # predecessor.
+                $v_->{pred} = $edge; 
+            }
+        }
+    }
+
+
+    # Retrieve the shortest path from "start" to $targetPath.
+    my @path = ();
+    my $cur = $targetPath;
+    return () unless defined $graph{$targetPath}->{pred};
+    while ($cur ne "start") {
+        push @path, $graph{$cur}->{pred};
+        $cur = $graph{$cur}->{pred}->{start};
+    }
+
+    return @path;
+}
+
+
 # Parse the arguments.
 
 if ($ARGV[0] eq "--query") {
@@ -46,6 +195,7 @@ if ($ARGV[0] eq "--query") {
 
         elsif ($cmd eq "info") {
             my $storePath = <STDIN>; chomp $storePath;
+
             my $info;
             if (defined $narFiles{$storePath}) {
                 $info = @{$narFiles{$storePath}}[0];
@@ -57,13 +207,32 @@ if ($ARGV[0] eq "--query") {
                 print "0\n";
                 next; # not an error
             }
+
             print "1\n";
             print "$info->{deriver}\n";
             my @references = split " ", $info->{references};
             print scalar @references, "\n";
             print "$_\n" foreach @references;
-            my $size = $info->{size} || 0;
-            print "$size\n";
+
+            my @path = computeSmallestDownload $storePath;
+            
+            my $downloadSize = 0;
+            while (scalar @path > 0) {
+                my $edge = pop @path;
+                my $u = $edge->{start};
+                my $v = $edge->{end};
+                if ($edge->{type} eq "patch") {
+                    $downloadSize += $edge->{info}->{size} || 0;
+                }
+                elsif ($edge->{type} eq "narfile") {
+                    $downloadSize += $edge->{info}->{size} || 0;
+                }
+            }
+                
+            print "$downloadSize\n";
+            
+            my $narSize = $info->{narSize} || 0;
+            print "$narSize\n";
         }
         
         else { die "unknown command `$cmd'"; }
@@ -79,6 +248,7 @@ elsif ($ARGV[0] ne "--substitute") {
 
 die unless scalar @ARGV == 2;
 my $targetPath = $ARGV[1];
+$fast = 0;
 
 
 # Create a temporary directory.
@@ -110,148 +280,9 @@ foreach my $localPath (@{$localPathList}) {
 }
 
 
-# Build a graph of all store paths that might contribute to the
-# construction of $targetPath, and the special node "start".  The
-# edges are either patch operations, or downloads of full NAR files.
-# The latter edges only occur between "start" and a store path.
-
-my %graph;
-
-$graph{"start"} = {d => 0, pred => undef, edges => []};
-
-my @queue = ();
-my $queueFront = 0;
-my %done;
-
-sub addToQueue {
-    my $v = shift;
-    return if defined $done{$v};
-    $done{$v} = 1;
-    push @queue, $v;
-}
-
-sub addNode {
-    my $u = shift;
-    $graph{$u} = {d => 999999999999, pred => undef, edges => []}
-        unless defined $graph{$u};
-}
-
-sub addEdge {
-    my $u = shift;
-    my $v = shift;
-    my $w = shift;
-    my $type = shift;
-    my $info = shift;
-    addNode $u;
-    push @{$graph{$u}->{edges}},
-        {weight => $w, start => $u, end => $v, type => $type, info => $info};
-    my $n = scalar @{$graph{$u}->{edges}};
-}
-
-addToQueue $targetPath;
-
-sub isValidPath {
-    my $p = shift;
-    return system("$binDir/nix-store --check-validity '$p' 2> /dev/null") == 0;
-}
-
-sub parseHash {
-    my $hash = shift;
-    if ($hash =~ /^(.+):(.+)$/) {
-        return ($1, $2);
-    } else {
-        return ("md5", $hash);
-    }
-}
-
-while ($queueFront < scalar @queue) {
-    my $u = $queue[$queueFront++];
-#    print "$u\n";
-
-    addNode $u;
-
-    # If the path already exists, it has distance 0 from the "start"
-    # node.
-    if (isValidPath($u)) {
-        addEdge "start", $u, 0, "present", undef;
-    }
-
-    else {
-
-        # Add patch edges.
-        my $patchList = $patches{$u};
-        foreach my $patch (@{$patchList}) {
-            if (isValidPath($patch->{basePath})) {
-                # !!! this should be cached
-                my ($baseHashAlgo, $baseHash) = parseHash $patch->{baseHash};
-                my $format = "--base32";
-                $format = "" if $baseHashAlgo eq "md5";
-                my $hash = `$binDir/nix-hash --type '$baseHashAlgo' $format "$patch->{basePath}"`;
-                chomp $hash;
-                if ($hash ne $baseHash) {
-                    print LOGFILE "$$ rejecting $patch->{basePath}\n";
-                    next;
-                }
-            }
-            addToQueue $patch->{basePath};
-            addEdge $patch->{basePath}, $u, $patch->{size}, "patch", $patch;
-        }
-
-        # Add NAR file edges to the start node.
-        my $narFileList = $narFiles{$u};
-        foreach my $narFile (@{$narFileList}) {
-            # !!! how to handle files whose size is not known in advance?
-            # For now, assume some arbitrary size (1 MB).
-            addEdge "start", $u, ($narFile->{size} || 1000000), "narfile", $narFile;
-            if ($u eq $targetPath) {
-                my $size = $narFile->{size} || -1;
-                print LOGFILE "$$ full-download-would-be $size\n";
-            }
-        }
-
-    }
-}
-
-
-# Run Dijkstra's shortest path algorithm to determine the shortest
-# sequence of download and/or patch actions that will produce
-# $targetPath.
-
-sub byDistance { # sort by distance, reversed
-    return -($graph{$a}->{d} <=> $graph{$b}->{d});
-}
-
-my @todo = keys %graph;
-
-while (scalar @todo > 0) {
-
-    # Remove the closest element from the todo list.
-    @todo = sort byDistance @todo;
-    my $u = pop @todo;
-
-    my $u_ = $graph{$u};
-
-    foreach my $edge (@{$u_->{edges}}) {
-        my $v_ = $graph{$edge->{end}};
-        if ($v_->{d} > $u_->{d} + $edge->{weight}) {
-            $v_->{d} = $u_->{d} + $edge->{weight};
-            # Store the edge; to edge->start is actually the
-            # predecessor.
-            $v_->{pred} = $edge; 
-        }
-    }
-}
-
-
-# Retrieve the shortest path from "start" to $targetPath.
-my @path = ();
-my $cur = $targetPath;
-die "don't know how to produce $targetPath\n"
-    unless defined $graph{$targetPath}->{pred};
-while ($cur ne "start") {
-    push @path, $graph{$cur}->{pred};
-    $cur = $graph{$cur}->{pred}->{start};
-}
+# Compute the shortest path.
+my @path = computeSmallestDownload $targetPath;
+die "don't know how to produce $targetPath\n" if scalar @path == 0;
 
 
 # Traverse the shortest path, perform the actions described by the

scripts/generate-patches.pl.in

@@ -1,408 +0,0 @@
-#! @perl@ -w -I@libexecdir@/nix
-
-use strict;
-use File::Temp qw(tempdir);
-use readmanifest;
-
-
-# Some patch generations options.
-
-# Max size of NAR archives to generate patches for.
-my $maxNarSize = $ENV{"NIX_MAX_NAR_SIZE"};
-$maxNarSize = 100 * 1024 * 1024 if !defined $maxNarSize;
-
-# If patch is bigger than this fraction of full archive, reject.
-my $maxPatchFraction = $ENV{"NIX_PATCH_FRACTION"};
-$maxPatchFraction = 0.60 if !defined $maxPatchFraction;
-
-
-die unless scalar @ARGV == 5;
-
-my $hashAlgo = "sha256";
-
-my $cacheDir = $ARGV[0];
-my $patchesDir = $ARGV[1];
-my $patchesURL = $ARGV[2];
-my $srcDir = $ARGV[3];
-my $dstDir = $ARGV[4];
-
-my $tmpDir = tempdir("nix-generate-patches.XXXXXX", CLEANUP => 1, TMPDIR => 1)
-    or die "cannot create a temporary directory";
-
-print "TEMP = $tmpDir\n";
-
-#END { rmdir $tmpDir; }
-
-my %srcNarFiles;
-my %srcLocalPaths;
-my %srcPatches;
-
-my %dstNarFiles;
-my %dstLocalPaths;
-my %dstPatches;
-
-readManifest "$srcDir/MANIFEST",
-    \%srcNarFiles, \%srcLocalPaths, \%srcPatches;
-
-readManifest "$dstDir/MANIFEST",
-    \%dstNarFiles, \%dstLocalPaths, \%dstPatches;
-
-
-sub findOutputPaths {
-    my $narFiles = shift;
-
-    my %outPaths;
-    
-    foreach my $p (keys %{$narFiles}) {
-
-        # Ignore store expressions.
-        next if ($p =~ /\.store$/);
-        next if ($p =~ /\.drv$/);
-        
-        # Ignore builders (too much ambiguity -- they're all called
-        # `builder.sh').
-        next if ($p =~ /\.sh$/);
-        next if ($p =~ /\.patch$/);
-        
-        # Don't bother including tar files etc.
-        next if ($p =~ /\.tar\.(gz|bz2)$/ || $p =~ /\.zip$/ || $p =~ /\.bin$/);
-
-        $outPaths{$p} = 1;
-    }
-
-    return %outPaths;
-}
-
-print "finding src output paths...\n";
-my %srcOutPaths = findOutputPaths \%srcNarFiles;
-
-print "finding dst output paths...\n";
-my %dstOutPaths = findOutputPaths \%dstNarFiles;
-
-
-sub getNameVersion {
-    my $p = shift;
-    $p =~ /\/[0-9a-z]+((?:-[a-zA-Z][^\/-]*)+)([^\/]*)$/;
-    my $name = $1;
-    my $version = $2;
-    $name =~ s/^-//;
-    $version =~ s/^-//;
-    return ($name, $version);
-}
-
-
-# A quick hack to get a measure of the `distance' between two
-# versions: it's just the position of the first character that differs
-# (or 999 if they are the same).
-sub versionDiff {
-    my $s = shift;
-    my $t = shift;
-    my $i;
-    return 999 if $s eq $t;
-    for ($i = 0; $i < length $s; $i++) {
-        return $i if $i >= length $t or
-            substr($s, $i, 1) ne substr($t, $i, 1);
-    }
-    return $i;
-}
-
-
-sub getNarBz2 {
-    my $narFiles = shift;
-    my $storePath = shift;
-    
-    my $narFileList = $$narFiles{$storePath};
-    die "missing store expression $storePath" unless defined $narFileList;
-
-    my $narFile = @{$narFileList}[0];
-    die unless defined $narFile;
-
-    $narFile->{url} =~ /\/([^\/]+)$/;
-    die unless defined $1;
-    return "$cacheDir/$1";
-}
-
-
-sub containsPatch {
-    my $patches = shift;
-    my $storePath = shift;
-    my $basePath = shift;
-    my $patchList = $$patches{$storePath};
-    return 0 if !defined $patchList;
-    my $found = 0;
-    foreach my $patch (@{$patchList}) {
-        # !!! baseHash might differ
-        return 1 if $patch->{basePath} eq $basePath;
-    }
-    return 0;
-}
-
-
-# Compute the "weighted" number of uses of a path in the build graph.
-sub computeUses {
-    my $narFiles = shift;
-    my $path = shift;
-
-    # Find the deriver of $path.
-    return 1 unless defined $$narFiles{$path};
-    my $deriver = @{$$narFiles{$path}}[0]->{deriver};
-    return 1 unless defined $deriver && $deriver ne "";
-
-#    print "  DERIVER $deriver\n";
-
-    # Optimisation: build the referrers graph from the references
-    # graph.
-    my %referrers;
-    foreach my $q (keys %{$narFiles}) {
-        my @refs = split " ", @{$$narFiles{$q}}[0]->{references};
-        foreach my $r (@refs) {
-            $referrers{$r} = [] unless defined $referrers{$r};
-            push @{$referrers{$r}}, $q;
-        }
-    }
-
-    # Determine the shortest path from $deriver to all other reachable
-    # paths in the `referrers' graph.
-
-    my %dist;
-    $dist{$deriver} = 0;
-
-    my @queue = ($deriver);
-    my $pos = 0;
-    
-    while ($pos < scalar @queue) {
-        my $p = $queue[$pos];
-        $pos++;
-
-        foreach my $q (@{$referrers{$p}}) {
-            if (!defined $dist{$q}) {
-                $dist{$q} = $dist{$p} + 1;
-#                print "    $q $dist{$q}\n";
-                push @queue, $q;
-            }
-        }
-    }
-
-    my $wuse = 1.0;
-    foreach my $user (keys %dist) {
-        next if $user eq $deriver;
-#        print "    $user $dist{$user}\n";
-        $wuse += 1.0 / 2.0**$dist{$user};
-    }
-
-#    print "    XXX $path $wuse\n";
-    
-    return $wuse;
-}
-
-
-# For each output path in the destination, see if we need to / can
-# create a patch.
-
-print "creating patches...\n";
-
-foreach my $p (keys %dstOutPaths) {
-
-    # If exactly the same path already exists in the source, skip it.
-    next if defined $srcOutPaths{$p};
-    
-    print "  $p\n";
-
-    # If not, then we should find the paths in the source that are
-    # `most' likely to be present on a system that wants to install
-    # this path.
-
-    (my $name, my $version) = getNameVersion $p;
-
-    my @closest = ();
-    my $closestVersion;
-    my $minDist = -1; # actually, larger means closer
-
-    # Find all source paths with the same name.
-
-    foreach my $q (keys %srcOutPaths) {
-        (my $name2, my $version2) = getNameVersion $q;
-        if ($name eq $name2) {
-
-            # If the sizes differ too much, then skip.  This
-            # disambiguates between, e.g., a real component and a
-            # wrapper component (cf. Firefox in Nixpkgs).
-            my $srcSize = @{$srcNarFiles{$q}}[0]->{size};
-            my $dstSize = @{$dstNarFiles{$p}}[0]->{size};
-            my $ratio = $srcSize / $dstSize;
-            $ratio = 1 / $ratio if $ratio < 1;
-#            print "  SIZE $srcSize $dstSize $ratio $q\n";
-
-            if ($ratio >= 3) {
-                print "  SKIPPING $q due to size ratio $ratio ($srcSize $dstSize)\n";
-                next;
-            }
-
-            # If the numbers of weighted uses differ too much, then
-            # skip.  This disambiguates between, e.g., the bootstrap
-            # GCC and the final GCC in Nixpkgs.
-            my $srcUses = computeUses \%srcNarFiles, $q;
-            my $dstUses = computeUses \%dstNarFiles, $p;
-            $ratio = $srcUses / $dstUses;
-            $ratio = 1 / $ratio if $ratio < 1;
-            print "  USE $srcUses $dstUses $ratio $q\n";
-            
-#            if ($ratio >= 2) {
-#                print "  SKIPPING $q due to use ratio $ratio ($srcUses $dstUses)\n";
-#                next;
-#            }
-
-            # If there are multiple matching names, include the ones
-            # with the closest version numbers.
-            my $dist = versionDiff $version, $version2;
-            if ($dist > $minDist) {
-                $minDist = $dist;
-                @closest = ($q);
-                $closestVersion = $version2;
-            } elsif ($dist == $minDist) {
-                push @closest, $q;
-            }
-        }
-    }
-
-    if (scalar(@closest) == 0) {
-        print "  NO BASE: $p\n";
-        next;
-    }
-
-    foreach my $closest (@closest) {
-
-        # Generate a patch between $closest and $p.
-        print "  $p <- $closest\n";
-
-        # If the patch already exists, skip it.
-        if (containsPatch(\%srcPatches, $p, $closest) ||
-            containsPatch(\%dstPatches, $p, $closest))
-        {
-            print "    skipping, already exists\n";
-            next;
-        }
-
-#        next;
-        
-        my $srcNarBz2 = getNarBz2 \%srcNarFiles, $closest;
-        my $dstNarBz2 = getNarBz2 \%dstNarFiles, $p;
-
-        system("@bunzip2@ < $srcNarBz2 > $tmpDir/A") == 0
-            or die "cannot unpack $srcNarBz2";
-
-        if ((stat "$tmpDir/A")[7] >= $maxNarSize) {
-            print "    skipping, source is too large\n";
-            next;
-        }
-        
-        system("@bunzip2@ < $dstNarBz2 > $tmpDir/B") == 0
-            or die "cannot unpack $dstNarBz2";
-
-        if ((stat "$tmpDir/B")[7] >= $maxNarSize) {
-            print "    skipping, destination is too large\n";
-            next;
-        }
-        
-        system("@libexecdir@/bsdiff $tmpDir/A $tmpDir/B $tmpDir/DIFF") == 0
-            or die "cannot compute binary diff";
-
-        my $baseHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/A` or die;
-        chomp $baseHash;
-
-        my $narHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/B` or die;
-        chomp $narHash;
-
-        my $narDiffHash = `@bindir@/nix-hash --flat --type $hashAlgo --base32 $tmpDir/DIFF` or die;
-        chomp $narDiffHash;
-
-        my $narDiffSize = (stat "$tmpDir/DIFF")[7];
-        my $dstNarBz2Size = (stat $dstNarBz2)[7];
-
-        print "    size $narDiffSize; full size $dstNarBz2Size\n";
-        
-        if ($narDiffSize >= $dstNarBz2Size) {
-            print "    rejecting; patch bigger than full archive\n";
-            next;
-        }
-    
-        if ($narDiffSize / $dstNarBz2Size >= $maxPatchFraction) {
-            print "    rejecting; patch too large relative to full archive\n";
-            next;
-        }
-    
-        my $finalName =
-            "$narDiffHash.nar-bsdiff";
-
-        if (-e "$patchesDir/$finalName") {
-            print "    not copying, already exists\n";
-        }
-
-        else {
-
-            system("cp '$tmpDir/DIFF' '$patchesDir/$finalName.tmp'") == 0
-                or die "cannot copy diff";
-            
-            rename("$patchesDir/$finalName.tmp", "$patchesDir/$finalName")
-                or die "cannot rename $patchesDir/$finalName.tmp";
-            
-        }
-        
-        # Add the patch to the manifest.
-        addPatch \%dstPatches, $p,
-            { url => "$patchesURL/$finalName", hash => "$hashAlgo:$narDiffHash"
-            , size => $narDiffSize, basePath => $closest, baseHash => "$hashAlgo:$baseHash"
-            , narHash => "$hashAlgo:$narHash", patchType => "nar-bsdiff"
-            }, 0;
-    }
-}
-
-
-# Add in any potentially useful patches in the source (namely, those
-# patches that produce either paths in the destination or paths that
-# can be used as the base for other useful patches).
-
-print "propagating patches...\n";
-
-my $changed;
-do {
-    # !!! we repeat this to reach the transitive closure; inefficient
-    $changed = 0;
-
-    print "loop\n";
-
-    my %dstBasePaths;
-    foreach my $q (keys %dstPatches) {
-	foreach my $patch (@{$dstPatches{$q}}) {
-	    $dstBasePaths{$patch->{basePath}} = 1;
-	}
-    }
-
-    foreach my $p (keys %srcPatches) {
-        my $patchList = $srcPatches{$p};
-
-        my $include = 0;
-
-        # Is path $p included in the destination?  If so, include
-        # patches that produce it.
-        $include = 1 if defined $dstNarFiles{$p};
-
-        # Is path $p a path that serves as a base for paths in the
-        # destination?  If so, include patches that produce it.
-	# !!! check baseHash
-        $include = 1 if defined $dstBasePaths{$p};
-
-        if ($include) {
-            foreach my $patch (@{$patchList}) {
-                $changed = 1 if addPatch \%dstPatches, $p, $patch;
-            }
-        }
-        
-    }
-    
-} while $changed;
-
-
-# Rewrite the manifest of the destination (with the new patches).
-writeManifest "$dstDir/MANIFEST",
-    \%dstNarFiles, \%dstPatches;

scripts/nix-build.in

@@ -102,7 +102,7 @@ EOF
         $n += 2;
     }
 
-    elsif ($arg eq "--max-jobs" or $arg eq "-j" or $arg eq "--max-silent-time" or $arg eq "--log-type") {
+    elsif ($arg eq "--max-jobs" or $arg eq "-j" or $arg eq "--max-silent-time" or $arg eq "--log-type" or $arg eq "--cores") {
         $n++;
         die "$0: `$arg' requires an argument\n" unless $n < scalar @ARGV;
         push @buildArgs, ($arg, $ARGV[$n]);
@@ -123,6 +123,11 @@ EOF
         $verbose = 1;
     }
     
+    elsif ($arg eq "--quiet") {
+        push @buildArgs, $arg;
+        push @instArgs, $arg;
+    }
+    
     elsif (substr($arg, 0, 1) eq "-") {
         push @buildArgs, $arg;
     }
@@ -165,7 +170,7 @@ foreach my $expr (@exprs) {
 
     # Build.
     my @outPaths;
-    $pid = open(OUTPATHS, "-|") || exec "$binDir/nix-store", "--add-root", $outLink, "--indirect", "-rv",
+    $pid = open(OUTPATHS, "-|") || exec "$binDir/nix-store", "--add-root", $outLink, "--indirect", "-r",
         @buildArgs, @drvPaths;
     while (<OUTPATHS>) {chomp; push @outPaths, $_;}
     if (!close OUTPATHS) {

scripts/nix-copy-closure.in

@@ -1,6 +1,6 @@
 #! @perl@ -w -I@libexecdir@/nix
 
-use ssh;
+use SSH;
 
 my $binDir = $ENV{"NIX_BIN_DIR"} || "@bindir@";
 
@@ -61,7 +61,7 @@ if ($toMode) { # Copy TO the remote machine.
     my @allStorePaths;
 
     # Get the closure of this path.
-    my $pid = open(READ, "$binDir/nix-store --query --requisites @storePaths|") or die;
+    my $pid = open(READ, "set -f; $binDir/nix-store --query --requisites @storePaths|") or die;
     
     while (<READ>) {
         chomp;
@@ -73,7 +73,7 @@ if ($toMode) { # Copy TO the remote machine.
 
 
     # Ask the remote host which paths are invalid.
-    open(READ, "ssh $sshHost @sshOpts nix-store --check-validity --print-invalid @allStorePaths|");
+    open(READ, "set -f; ssh $sshHost @sshOpts nix-store --check-validity --print-invalid @allStorePaths|");
     my @missing = ();
     while (<READ>) {
         chomp;
@@ -88,7 +88,7 @@ if ($toMode) { # Copy TO the remote machine.
         print STDERR "  $_\n" foreach @missing;
         my $extraOpts = "";
         $extraOpts .= "--sign" if $sign == 1;
-        system("nix-store --export $extraOpts @missing $compressor | ssh $sshHost @sshOpts '$decompressor nix-store --import'") == 0
+        system("set -f; nix-store --export $extraOpts @missing $compressor | ssh $sshHost @sshOpts '$decompressor nix-store --import'") == 0
             or die "copying store paths to remote machine `$sshHost' failed: $?";
     }
 
@@ -101,7 +101,7 @@ else { # Copy FROM the remote machine.
     # machine.  Paths are assumed to be store paths; there is no
     # resolution (following of symlinks).
     my $pid = open(READ,
-        "ssh @sshOpts $sshHost nix-store --query --requisites @storePaths|") or die;
+        "set -f; ssh @sshOpts $sshHost nix-store --query --requisites @storePaths|") or die;
     
     my @allStorePaths;
 
@@ -115,7 +115,7 @@ else { # Copy FROM the remote machine.
 
 
     # What paths are already valid locally?
-    open(READ, "@bindir@/nix-store --check-validity --print-invalid @allStorePaths|");
+    open(READ, "set -f; @bindir@/nix-store --check-validity --print-invalid @allStorePaths|");
     my @missing = ();
     while (<READ>) {
         chomp;
@@ -130,7 +130,7 @@ else { # Copy FROM the remote machine.
         print STDERR "  $_\n" foreach @missing;
         my $extraOpts = "";
         $extraOpts .= "--sign" if $sign == 1;
-        system("ssh $sshHost @sshOpts 'nix-store --export $extraOpts @missing $compressor' | $decompressor @bindir@/nix-store --import") == 0
+        system("set -f; ssh $sshHost @sshOpts 'nix-store --export $extraOpts @missing $compressor' | $decompressor @bindir@/nix-store --import") == 0
             or die "copying store paths from remote machine `$sshHost' failed: $?";
     }
 

scripts/nix-generate-patches.in

@@ -0,0 +1,42 @@
+#! @perl@ -w -I@libexecdir@/nix
+
+use strict;
+use File::Temp qw(tempdir);
+use NixManifest;
+use GeneratePatches;
+
+if (scalar @ARGV != 5) {
+    print STDERR <<EOF;
+Usage: nix-generate-patches NAR-DIR PATCH-DIR PATCH-URI OLD-MANIFEST NEW-MANIFEST
+
+This command generates binary patches between NAR files listed in
+OLD-MANIFEST and NEW-MANIFEST.  The patches are written to the
+directory PATCH-DIR, and the prefix PATCH-URI is used to generate URIs
+for the patches.  The patches are added to NEW-MANIFEST.  All NARs are
+required to exist in NAR-DIR.  Patches are generated between
+succeeding versions of packages with the same name.
+EOF
+    exit 1;
+}
+
+my $narPath = $ARGV[0];
+my $patchesPath = $ARGV[1];
+my $patchesURL = $ARGV[2];
+my $srcManifest = $ARGV[3];
+my $dstManifest = $ARGV[4];
+
+my (%srcNarFiles, %srcLocalPaths, %srcPatches);
+readManifest $srcManifest, \%srcNarFiles, \%srcLocalPaths, \%srcPatches;
+
+my (%dstNarFiles, %dstLocalPaths, %dstPatches);
+readManifest $dstManifest, \%dstNarFiles, \%dstLocalPaths, \%dstPatches;
+
+my $tmpDir = tempdir("nix-generate-patches.XXXXXX", CLEANUP => 1, TMPDIR => 1)
+    or die "cannot create a temporary directory";
+
+generatePatches \%srcNarFiles, \%dstNarFiles, \%srcPatches, \%dstPatches,
+    $narPath, $patchesPath, $patchesURL, $tmpDir;
+
+propagatePatches \%srcPatches, \%dstNarFiles, \%dstPatches;
+
+writeManifest $dstManifest, \%dstNarFiles, \%dstPatches;

scripts/nix-prefetch-url.in

@@ -21,7 +21,10 @@ if test -z "$url"; then
     exit 1
 fi
 
-name=$(basename "$url")
+# Handle escaped characters in the URI.  `+', `=' and `?' are the only
+# characters that are valid in Nix store path names but have a special
+# meaning in URIs.
+name=$(basename "$url" | @sed@ -e 's/%2b/+/g' -e 's/%3d/=/g' -e 's/%3f/\?/g')
 if test -z "$name"; then echo "invalid url"; exit 1; fi
 
 

scripts/nix-pull.in

@@ -2,7 +2,7 @@
 
 use strict;
 use File::Temp qw(tempdir);
-use readmanifest;
+use NixManifest;
 
 my $tmpDir = tempdir("nix-pull.XXXXXX", CLEANUP => 1, TMPDIR => 1)
     or die "cannot create a temporary directory";

scripts/nix-push.in

@@ -2,7 +2,7 @@
 
 use strict;
 use File::Temp qw(tempdir);
-use readmanifest;
+use NixManifest;
 
 my $hashAlgo = "sha256";
 
@@ -172,12 +172,6 @@ for (my $n = 0; $n < scalar @storePaths; $n++) {
     $narbz2Hash =~ /^[0-9a-z]+$/ or die "invalid hash";
     close HASH;
 
-    open HASH, "$narDir/nar-hash" or die "cannot open nar-hash";
-    my $narHash = <HASH>;
-    chomp $narHash;
-    $narHash =~ /^[0-9a-z]+$/ or die "invalid hash";
-    close HASH;
-    
     my $narName = "$narbz2Hash.nar.bz2";
 
     my $narFile = "$narDir/$narName";
@@ -195,6 +189,14 @@ for (my $n = 0; $n < scalar @storePaths; $n++) {
     chomp $deriver;
     $deriver = "" if $deriver eq "unknown-deriver";
 
+    my $narHash = `$binDir/nix-store --query --hash '$storePath'`;
+    die "cannot query hash for `$storePath'" if $? != 0;
+    chomp $narHash;
+
+    my $narSize = `$binDir/nix-store --query --size '$storePath'`;
+    die "cannot query size for `$storePath'" if $? != 0;
+    chomp $narSize;
+
     my $url;
     if ($localCopy) {
         $url = "$targetArchivesUrl/$narName";
@@ -205,7 +207,8 @@ for (my $n = 0; $n < scalar @storePaths; $n++) {
         { url => $url
         , hash => "$hashAlgo:$narbz2Hash"
         , size => $narbz2Size
-        , narHash => "$hashAlgo:$narHash"
+        , narHash => "$narHash"
+        , narSize => $narSize
         , references => $references
         , deriver => $deriver
         }

src/Makefile.am

@@ -1,5 +1,3 @@
 SUBDIRS = bin2c boost libutil libstore libmain nix-store nix-hash \
  libexpr nix-instantiate nix-env nix-worker nix-setuid-helper \
  nix-log2xml bsdiff-4.3
-
-EXTRA_DIST = aterm-helper.pl

src/aterm-helper.pl

@@ -1,178 +0,0 @@
-#! /usr/bin/perl -w
-
-# This program generates C/C++ code for efficiently manipulating
-# ATerms.  It generates functions to build and match ATerms according
-# to a set of constructor definitions defined in a file read from
-# standard input.  A constructor is defined by a line with the
-# following format:
-#
-#   SYM | ARGS | TYPE | FUN?
-#
-# where SYM is the name of the constructor, ARGS is a
-# whitespace-separated list of argument types, TYPE is the type of the
-# resulting ATerm (which should be `ATerm' or a type synonym for
-# `ATerm'), and the optional FUN is used to construct the names of the
-# build and match functions (it defaults to SYM; overriding it is
-# useful if there are overloaded constructors, e.g., with different
-# arities).  Note that SYM may be empty.
-#
-# A line of the form
-#
-#   VAR = EXPR
-#
-# causes a ATerm variable to be generated that is initialised to the
-# value EXPR.
-#
-# Finally, a line of the form
-#
-#   init NAME
-#
-# causes the initialisation function to be called `NAME'.  This
-# function must be called before any of the build/match functions or
-# the generated variables are used.
-
-die if scalar @ARGV != 2;
-
-my $syms = "";
-my $init = "";
-my $initFun = "init";
-
-open HEADER, ">$ARGV[0]";
-open IMPL, ">$ARGV[1]";
-
-print HEADER "#include <aterm2.h>\n";
-print HEADER "#ifdef __cplusplus\n";
-print HEADER "namespace nix {\n";
-print HEADER "#endif\n\n\n";
-print IMPL "namespace nix {\n";
-
-while (<STDIN>) {
-    s/\#.*//;
-    next if (/^\s*$/);
-    
-    if (/^\s*(\w*)\s*\|([^\|]*)\|\s*(\w+)\s*\|\s*(\w+)?/) {
-        my $const = $1;
-        my @types = split ' ', $2;
-        my $result = $3;
-        my $funname = $4;
-        $funname = $const unless defined $funname;
-
-        my $formals = "";
-        my $formals2 = "";
-        my $args = "";
-        my $unpack = "";
-        my $n = 1;
-        foreach my $type (@types) {
-            my $realType = $type;
-            $args .= ", ";
-            if ($type eq "string") {
-#                $args .= "(ATerm) ATmakeAppl0(ATmakeAFun((char *) e$n, 0, ATtrue))";
-#                $type = "const char *";
-                $type = "ATerm";
-                $args .= "e$n";
-                # !!! in the matcher, we should check that the
-                # argument is a string (i.e., a nullary application).
-            } elsif ($type eq "int") {
-                $args .= "(ATerm) ATmakeInt(e$n)";
-            } elsif ($type eq "ATermList" || $type eq "ATermBlob") {
-                $args .= "(ATerm) e$n";
-            } else {
-                $args .= "e$n";
-            }
-            $formals .= ", " if $formals ne "";
-            $formals .= "$type e$n";
-            $formals2 .= ", ";
-            $formals2 .= "$type & e$n";
-            my $m = $n - 1;
-            # !!! more checks here
-            if ($type eq "int") {
-                $unpack .= "    e$n = ATgetInt((ATermInt) ATgetArgument(e, $m));\n";
-            } elsif ($type eq "ATermList") {
-                $unpack .= "    e$n = (ATermList) ATgetArgument(e, $m);\n";
-            } elsif ($type eq "ATermBlob") {
-                $unpack .= "    e$n = (ATermBlob) ATgetArgument(e, $m);\n";
-            } elsif ($realType eq "string") {
-                $unpack .= "    e$n = ATgetArgument(e, $m);\n";
-                $unpack .= "    if (ATgetType(e$n) != AT_APPL) return false;\n";
-            } else {
-                $unpack .= "    e$n = ATgetArgument(e, $m);\n";
-            }
-            $n++;
-        }
-
-        my $arity = scalar @types;
-
-        print HEADER "extern AFun sym$funname;\n\n";
-        
-        print IMPL "AFun sym$funname = 0;\n";
-        
-        if ($arity == 0) {
-            print HEADER "extern ATerm const$funname;\n\n";
-            print IMPL "ATerm const$funname = 0;\n";
-        }
-        
-        print HEADER "static inline $result make$funname($formals) __attribute__ ((pure, nothrow));\n";
-        print HEADER "static inline $result make$funname($formals) {\n";
-        if ($arity == 0) {
-            print HEADER "    return const$funname;\n";
-        }
-        elsif ($arity <= 6) {
-            print HEADER "    return (ATerm) ATmakeAppl$arity(sym$funname$args);\n";
-        } else {
-            $args =~ s/^,//;
-            print HEADER "    ATerm array[$arity] = {$args};\n";
-            print HEADER "    return (ATerm) ATmakeApplArray(sym$funname, array);\n";
-        }
-        print HEADER "}\n\n";
-
-        print HEADER "#ifdef __cplusplus\n";
-        print HEADER "static inline bool match$funname(ATerm e$formals2) {\n";
-        print HEADER "    if (ATgetType(e) != AT_APPL || (AFun) ATgetAFun(e) != sym$funname) return false;\n";
-        print HEADER "$unpack";
-        print HEADER "    return true;\n";
-        print HEADER "}\n";
-        print HEADER "#endif\n\n\n";
-
-        $init .= "    sym$funname = ATmakeAFun(\"$const\", $arity, ATfalse);\n";
-        $init .= "    ATprotectAFun(sym$funname);\n";
-        if ($arity == 0) {
-            $init .= "    const$funname = (ATerm) ATmakeAppl0(sym$funname);\n";
-            $init .= "    ATprotect(&const$funname);\n";
-        }
-    }
-
-    elsif (/^\s*(\w+)\s*=\s*(.*)$/) {
-        my $name = $1;
-        my $value = $2;
-        print HEADER "extern ATerm $name;\n";
-        print IMPL "ATerm $name = 0;\n";
-        $init .= "    $name = $value;\n";
-    }
-
-    elsif (/^\s*init\s+(\w+)\s*$/) {
-        $initFun = $1;
-    }
-
-    else {
-        die "bad line: `$_'";
-    }
-}
-
-print HEADER "void $initFun();\n\n";
-
-print HEADER "static inline const char * aterm2String(ATerm t) {\n";
-print HEADER "    return (const char *) ATgetName(ATgetAFun(t));\n";
-print HEADER "}\n\n";
-
-print IMPL "\n";
-print IMPL "void $initFun() {\n";
-print IMPL "$init";
-print IMPL "}\n";
-
-print HEADER "#ifdef __cplusplus\n";
-print HEADER "}\n";
-print HEADER "#endif\n\n\n";
-print IMPL "}\n";
-
-close HEADER;
-close IMPL;

src/bin2c/Makefile.am

@@ -1,3 +1,6 @@
 noinst_PROGRAMS = bin2c
 
 bin2c_SOURCES = bin2c.c
+
+bin2c$(EXEEXT): bin2c.c
+	$(CC_FOR_BUILD) $(CFLAGS_FOR_BUILD) -o bin2c bin2c.c

src/bin2c/bin2c.c

@@ -14,10 +14,10 @@ int main(int argc, char * * argv)
 {
     int c;
     if (argc != 2) abort();
-    print("static unsigned char %s[] = {", argv[1]);
+    print("static unsigned char %s[] = { ", argv[1]);
     while ((c = getchar()) != EOF) {
         print("0x%02x, ", (unsigned char) c);
     }
-    print("};\n");
+    print("0 };\n");
     return 0;
 }

src/bsdiff-4.3/Makefile.am

@@ -1,3 +1,5 @@
+EXTRA_DIST = compat-include
+
 libexec_PROGRAMS = bsdiff bspatch
 
 bsdiff_SOURCES = bsdiff.c
@@ -8,4 +10,4 @@ bspatch_SOURCES = bspatch.c
 
 bspatch_LDADD = ${bzip2_lib}
 
-AM_CFLAGS = -O3 ${bzip2_include}
+AM_CFLAGS = -O3 ${bzip2_include} ${bsddiff_compat_include}

src/bsdiff-4.3/bsdiff.c

@@ -277,6 +277,7 @@ int main(int argc,char *argv[])
 		for(scsc=scan+=len;scan<newsize;scan++) {
 			len=search(I,old,oldsize,new+scan,newsize-scan,
 					0,oldsize,&pos);
+			if (len > 64 * 1024) break;
 
 			for(;scsc<scan+len;scsc++)
 			if((scsc+lastoffset<oldsize) &&

src/bsdiff-4.3/compat-include/err.h

@@ -0,0 +1,12 @@
+/* Simulate BSD's <err.h> functionality. */
+
+#ifndef COMPAT_ERR_H_INCLUDED
+#define COMPAT_ERR_H_INCLUDED 1
+
+#include <stdio.h>
+#include <stdlib.h>
+
+#define err(rc,...)  do { fprintf(stderr,__VA_ARGS__); exit(rc); } while(0)
+#define errx(rc,...) do { fprintf(stderr,__VA_ARGS__); exit(rc); } while(0)
+
+#endif

src/libexpr/Makefile.am

@@ -2,27 +2,25 @@ pkglib_LTLIBRARIES = libexpr.la
 
 libexpr_la_SOURCES = \
  nixexpr.cc eval.cc primops.cc lexer-tab.cc parser-tab.cc \
- get-drvs.cc attr-path.cc expr-to-xml.cc common-opts.cc \
+ get-drvs.cc attr-path.cc value-to-xml.cc common-opts.cc \
  names.cc
 
 pkginclude_HEADERS = \
  nixexpr.hh eval.hh parser.hh lexer-tab.hh parser-tab.hh \
- get-drvs.hh attr-path.hh expr-to-xml.hh common-opts.hh \
- names.hh nixexpr-ast.hh
+ get-drvs.hh attr-path.hh value-to-xml.hh common-opts.hh \
+ names.hh symbol-table.hh
 
 libexpr_la_LIBADD = ../libutil/libutil.la ../libstore/libstore.la \
- ../boost/format/libformat.la
+ ../boost/format/libformat.la @boehmgc_lib@
 
-BUILT_SOURCES = nixexpr-ast.cc nixexpr-ast.hh \
+BUILT_SOURCES = \
  parser-tab.hh lexer-tab.hh parser-tab.cc lexer-tab.cc
 
-EXTRA_DIST = lexer.l parser.y nixexpr-ast.def nixexpr-ast.cc
+EXTRA_DIST = lexer.l parser.y
 
 AM_CXXFLAGS = \
- -I$(srcdir)/.. ${aterm_include} \
+ -I$(srcdir)/.. \
  -I$(srcdir)/../libutil -I$(srcdir)/../libstore
-AM_CFLAGS = \
- ${aterm_include}
 
 
 # Parser generation.
@@ -34,15 +32,6 @@ lexer-tab.cc lexer-tab.hh: lexer.l
 	$(flex) --outfile lexer-tab.cc --header-file=lexer-tab.hh $(srcdir)/lexer.l 
 
 
-# ATerm helper function generation.
-
-nixexpr-ast.cc nixexpr-ast.hh: ../aterm-helper.pl nixexpr-ast.def
-	$(perl) $(srcdir)/../aterm-helper.pl nixexpr-ast.hh nixexpr-ast.cc < $(srcdir)/nixexpr-ast.def
-
-
-CLEANFILES =
-
-
 # SDF stuff (not built by default).
 nix.tbl: nix.sdf
 	sdf2table -m Nix -s -i nix.sdf -o nix.tbl

src/libexpr/attr-path.cc

@@ -1,23 +1,13 @@
 #include "attr-path.hh"
-#include "nixexpr-ast.hh"
 #include "util.hh"
 
 
 namespace nix {
 
 
-bool isAttrs(EvalState & state, Expr e, ATermMap & attrs)
-{
-    e = evalExpr(state, e);
-    ATermList dummy;
-    if (!matchAttrs(e, dummy)) return false;
-    queryAllAttrs(e, attrs, false);
-    return true;
-}
-
-
-Expr findAlongAttrPath(EvalState & state, const string & attrPath,
-    const ATermMap & autoArgs, Expr e)
+// !!! Shouldn't we return a pointer to a Value?
+void findAlongAttrPath(EvalState & state, const string & attrPath,
+    Bindings & autoArgs, Expr * e, Value & v)
 {
     Strings tokens = tokenizeString(attrPath, ".");
 
@@ -25,8 +15,10 @@ Expr findAlongAttrPath(EvalState & state, const string & attrPath,
         Error(format("attribute selection path `%1%' does not match expression") % attrPath);
 
     string curPath;
+
+    state.mkThunk_(v, e);
     
-    for (Strings::iterator i = tokens.begin(); i != tokens.end(); ++i) {
+    foreach (Strings::iterator, i, tokens) {
 
         if (!curPath.empty()) curPath += ".";
         curPath += *i;
@@ -38,7 +30,10 @@ Expr findAlongAttrPath(EvalState & state, const string & attrPath,
         if (string2Int(attr, attrIndex)) apType = apIndex;
 
         /* Evaluate the expression. */
-        e = evalExpr(state, autoCallFunction(evalExpr(state, e), autoArgs));
+        Value vTmp;
+        state.autoCallFunction(autoArgs, v, vTmp);
+        v = vTmp;
+        state.forceValue(v);
 
         /* It should evaluate to either an attribute set or an
            expression, according to what is specified in the
@@ -46,36 +41,31 @@ Expr findAlongAttrPath(EvalState & state, const string & attrPath,
 
         if (apType == apAttr) {
 
-            ATermMap attrs;
-
-            if (!isAttrs(state, e, attrs))
+            if (v.type != tAttrs)
                 throw TypeError(
                     format("the expression selected by the selection path `%1%' should be an attribute set but is %2%")
-                    % curPath % showType(e));
-                
-            e = attrs.get(toATerm(attr));
-            if (!e)
-                throw Error(format("attribute `%1%' in selection path `%2%' not found") % attr % curPath);
+                    % curPath % showType(v));
 
+            Bindings::iterator a = v.attrs->find(state.symbols.create(attr));
+            if (a == v.attrs->end())
+                throw Error(format("attribute `%1%' in selection path `%2%' not found") % attr % curPath);
+            v = *a->value;
         }
 
         else if (apType == apIndex) {
 
-            ATermList es;
-            if (!matchList(e, es))
+            if (v.type != tList)
                 throw TypeError(
                     format("the expression selected by the selection path `%1%' should be a list but is %2%")
-                    % curPath % showType(e));
+                    % curPath % showType(v));
 
-            e = ATelementAt(es, attrIndex);
-            if (!e)
-                throw Error(format("list index %1% in selection path `%2%' not found") % attrIndex % curPath);
-            
+            if (attrIndex >= v.list.length)
+                throw Error(format("list index %1% in selection path `%2%' is out of range") % attrIndex % curPath);
+
+            v = *v.list.elems[attrIndex];
         }
         
     }
-    
-    return e;
 }
 
  

src/libexpr/attr-path.hh

@@ -1,17 +1,17 @@
 #ifndef __ATTR_PATH_H
 #define __ATTR_PATH_H
 
+#include "eval.hh"
+
 #include <string>
 #include <map>
 
-#include "eval.hh"
-
 
 namespace nix {
 
     
-Expr findAlongAttrPath(EvalState & state, const string & attrPath,
-    const ATermMap & autoArgs, Expr e);
+void findAlongAttrPath(EvalState & state, const string & attrPath,
+    Bindings & autoArgs, Expr * e, Value & v);
 
     
 }

src/libexpr/common-opts.cc

@@ -9,7 +9,7 @@ namespace nix {
 
 bool parseOptionArg(const string & arg, Strings::iterator & i,
     const Strings::iterator & argsEnd, EvalState & state,
-    ATermMap & autoArgs)
+    Bindings & autoArgs)
 {
     if (arg != "--arg" && arg != "--argstr") return false;
 
@@ -19,12 +19,18 @@ bool parseOptionArg(const string & arg, Strings::iterator & i,
     string name = *i++;
     if (i == argsEnd) throw error;
     string value = *i++;
-    
-    Expr e = arg == "--arg"
-        ? parseExprFromString(state, value, absPath("."))
-        : makeStr(value);
-    autoArgs.set(toATerm(name), e);
-    
+
+    /* !!! check for duplicates! */
+    Value * v = state.allocValue();
+    autoArgs.push_back(Attr(state.symbols.create(name), v));
+
+    if (arg == "--arg")
+        state.mkThunk_(*v, parseExprFromString(state, value, absPath(".")));
+    else
+        mkString(*v, value);
+
+    autoArgs.sort(); // !!! inefficient
+
     return true;
 }
 

src/libexpr/common-opts.hh

@@ -9,7 +9,7 @@ namespace nix {
 /* Some common option parsing between nix-env and nix-instantiate. */
 bool parseOptionArg(const string & arg, Strings::iterator & i,
     const Strings::iterator & argsEnd, EvalState & state,
-    ATermMap & autoArgs);
+    Bindings & autoArgs);
     
 }
 

src/libexpr/eval.hh

@@ -1,19 +1,193 @@
 #ifndef __EVAL_H
 #define __EVAL_H
 
+#include "nixexpr.hh"
+#include "symbol-table.hh"
+#include "hash.hh"
+
 #include <map>
 
-#include "aterm.hh"
-#include "nixexpr.hh"
+#if HAVE_BOEHMGC
+#include <gc/gc_allocator.h>
+#endif
 
 
 namespace nix {
 
 
-class Hash;
-    
+class EvalState;
+struct Env;
+struct Value;
+struct Attr;
+
+
+/* Attribute sets are represented as a vector of attributes, sorted by
+   symbol (i.e. pointer to the attribute name in the symbol table). */
+#if HAVE_BOEHMGC
+typedef std::vector<Attr, gc_allocator<Attr> > BindingsBase;
+#else
+typedef std::vector<Attr> BindingsBase;
+#endif
+
+
+class Bindings : public BindingsBase
+{
+public:
+    iterator find(const Symbol & name);
+    void sort();
+};
+
+
+typedef enum {
+    tInt = 1,
+    tBool,
+    tString,
+    tPath,
+    tNull,
+    tAttrs,
+    tList,
+    tThunk,
+    tApp,
+    tLambda,
+    tBlackhole,
+    tPrimOp,
+    tPrimOpApp,
+} ValueType;
+
+
+typedef void (* PrimOpFun) (EvalState & state, Value * * args, Value & v);
+
+
+struct PrimOp
+{
+    PrimOpFun fun;
+    unsigned int arity;
+    Symbol name;
+    PrimOp(PrimOpFun fun, unsigned int arity, Symbol name)
+        : fun(fun), arity(arity), name(name) { }
+};
+
+
+struct Value
+{
+    ValueType type;
+    union 
+    {
+        int integer;
+        bool boolean;
+        
+        /* Strings in the evaluator carry a so-called `context' (the
+           ATermList) which is a list of strings representing store
+           paths.  This is to allow users to write things like
+
+             "--with-freetype2-library=" + freetype + "/lib"
+
+           where `freetype' is a derivation (or a source to be copied
+           to the store).  If we just concatenated the strings without
+           keeping track of the referenced store paths, then if the
+           string is used as a derivation attribute, the derivation
+           will not have the correct dependencies in its inputDrvs and
+           inputSrcs.
+
+           The semantics of the context is as follows: when a string
+           with context C is used as a derivation attribute, then the
+           derivations in C will be added to the inputDrvs of the
+           derivation, and the other store paths in C will be added to
+           the inputSrcs of the derivations.
+
+           For canonicity, the store paths should be in sorted order. */
+        struct {
+            const char * s;
+            const char * * context; // must be in sorted order
+        } string;
+        
+        const char * path;
+        Bindings * attrs;
+        struct {
+            unsigned int length;
+            Value * * elems;
+        } list;
+        struct {
+            Env * env;
+            Expr * expr;
+        } thunk;
+        struct {
+            Value * left, * right;
+        } app;
+        struct {
+            Env * env;
+            ExprLambda * fun;
+        } lambda;
+        PrimOp * primOp;
+        struct {
+            Value * left, * right;
+        } primOpApp;
+    };
+};
+
+
+struct Env
+{
+    Env * up;
+    unsigned int prevWith; // nr of levels up to next `with' environment
+    Value * values[0];
+};
+
+
+struct Attr
+{
+    Symbol name;
+    Value * value;
+    Pos * pos;
+    Attr(Symbol name, Value * value, Pos * pos = &noPos)
+        : name(name), value(value), pos(pos) { };
+    Attr() : pos(&noPos) { };
+    bool operator < (const Attr & a) const
+    {
+        return name < a.name;
+    }
+};
+
+
+/* After overwriting an app node, be sure to clear pointers in the
+   Value to ensure that the target isn't kept alive unnecessarily. */
+static inline void clearValue(Value & v)
+{
+    v.app.right = 0;
+}
+
+
+static inline void mkInt(Value & v, int n)
+{
+    clearValue(v);
+    v.type = tInt;
+    v.integer = n;
+}
+
+
+static inline void mkBool(Value & v, bool b)
+{
+    clearValue(v);
+    v.type = tBool;
+    v.boolean = b;
+}
+
+
+static inline void mkApp(Value & v, Value & left, Value & right)
+{
+    v.type = tApp;
+    v.app.left = &left;
+    v.app.right = &right;
+}
+
+
+void mkString(Value & v, const char * s);
+void mkString(Value & v, const string & s, const PathSet & context = PathSet());
+void mkPath(Value & v, const char * s);
+
+void copyContext(const Value & v, PathSet & context);
+
 
-typedef std::map<Path, PathSet> DrvRoots;
 typedef std::map<Path, Hash> DrvHashes;
 
 /* Cache for calls to addToStore(); maps source paths to the store
@@ -22,75 +196,160 @@ typedef std::map<Path, Path> SrcToStore;
 
 struct EvalState;
 
-/* Note: using a ATermVector is safe here, since when we call a primop
-   we also have an ATermList on the stack. */
-typedef Expr (* PrimOp) (EvalState &, const ATermVector & args);
+
+std::ostream & operator << (std::ostream & str, const Value & v);
 
 
-struct EvalState 
+class EvalState 
 {
-    ATermMap normalForms;
-    ATermMap primOps;
-    DrvRoots drvRoots;
+public:
     DrvHashes drvHashes; /* normalised derivation hashes */
-    SrcToStore srcToStore; 
 
-    unsigned int nrEvaluated;
-    unsigned int nrCached;
+    SymbolTable symbols;
+
+    const Symbol sWith, sOutPath, sDrvPath, sType, sMeta, sName,
+        sSystem, sOverrides;
+
+private:
+    SrcToStore srcToStore; 
 
     bool allowUnsafeEquality;
 
-    EvalState();
+    std::map<Path, Expr *> parseTrees;
+
+public:
+    
+    EvalState();
+    ~EvalState();
+
+    /* Evaluate an expression read from the given file to normal
+       form. */
+    void evalFile(const Path & path, Value & v);
+
+    /* Evaluate an expression to normal form, storing the result in
+       value `v'. */
+    void eval(Expr * e, Value & v);
+    void eval(Env & env, Expr * e, Value & v);
+
+    /* Evaluation the expression, then verify that it has the expected
+       type. */
+    bool evalBool(Env & env, Expr * e);
+    void evalAttrs(Env & env, Expr * e, Value & v);
+
+    /* If `v' is a thunk, enter it and overwrite `v' with the result
+       of the evaluation of the thunk.  If `v' is a delayed function
+       application, call the function and overwrite `v' with the
+       result.  Otherwise, this is a no-op. */
+    void forceValue(Value & v);
+
+    /* Force a value, then recursively force list elements and
+       attributes. */
+    void strictForceValue(Value & v);
+
+    /* Force `v', and then verify that it has the expected type. */
+    int forceInt(Value & v);
+    bool forceBool(Value & v);
+    void forceAttrs(Value & v);
+    void forceList(Value & v);
+    void forceFunction(Value & v); // either lambda or primop
+    string forceString(Value & v);
+    string forceString(Value & v, PathSet & context);
+    string forceStringNoCtx(Value & v);
+
+    /* Return true iff the value `v' denotes a derivation (i.e. a
+       set with attribute `type = "derivation"'). */
+    bool isDerivation(Value & v);
+
+    /* String coercion.  Converts strings, paths and derivations to a
+       string.  If `coerceMore' is set, also converts nulls, integers,
+       booleans and lists to a string.  If `copyToStore' is set,
+       referenced paths are copied to the Nix store as a side effect.q */
+    string coerceToString(Value & v, PathSet & context,
+        bool coerceMore = false, bool copyToStore = true);
+
+    /* Path coercion.  Converts strings, paths and derivations to a
+       path.  The result is guaranteed to be a canonicalised, absolute
+       path.  Nothing is copied to the store. */
+    Path coerceToPath(Value & v, PathSet & context);
+
+private:
+
+    /* The base environment, containing the builtin functions and
+       values. */
+    Env & baseEnv;
+
+    unsigned int baseEnvDispl;
+
+public:
+    
+    /* The same, but used during parsing to resolve variables. */
+    StaticEnv staticBaseEnv; // !!! should be private
+
+private:
+    
+    void createBaseEnv();
+    
+    void addConstant(const string & name, Value & v);
 
-    void addPrimOps();
     void addPrimOp(const string & name,
-        unsigned int arity, PrimOp primOp);
+        unsigned int arity, PrimOpFun primOp);
+
+    Value * lookupVar(Env * env, const VarRef & var);
+    
+    friend class ExprVar;
+    friend class ExprAttrs;
+    friend class ExprLet;
+
+public:
+    
+    /* Do a deep equality test between two values.  That is, list
+       elements and attributes are compared recursively. */
+    bool eqValues(Value & v1, Value & v2);
+
+    void callFunction(Value & fun, Value & arg, Value & v);
+
+    /* Automatically call a function for which each argument has a
+       default value or has a binding in the `args' map. */
+    void autoCallFunction(Bindings & args, Value & fun, Value & res);
+    
+    /* Allocation primitives. */
+    Value * allocValue();
+    Env & allocEnv(unsigned int size);
+
+    Value * allocAttr(Value & vAttrs, const Symbol & name);
+
+    void mkList(Value & v, unsigned int length);
+    void mkAttrs(Value & v, unsigned int expected);
+    void mkThunk_(Value & v, Expr * expr);
+
+    Value * maybeThunk(Env & env, Expr * expr);
+    
+    /* Print statistics. */
+    void printStats();
+
+private:
+    
+    unsigned long nrEnvs;
+    unsigned long nrValuesInEnvs;
+    unsigned long nrValues;
+    unsigned long nrListElems;
+    unsigned long nrEvaluated;
+    unsigned long nrAttrsets;
+    unsigned long nrOpUpdates;
+    unsigned long nrOpUpdateValuesCopied;
+    unsigned int recursionDepth;
+    unsigned int maxRecursionDepth;
+    char * deepestStack; /* for measuring stack usage */
+    
+    friend class RecursionCounter;
+    friend class ExprOpUpdate;
 };
 
 
-/* Evaluate an expression to normal form. */
-Expr evalExpr(EvalState & state, Expr e);
+/* Return a string representing the type of the value `v'. */
+string showType(const Value & v);
 
-/* Evaluate an expression read from the given file to normal form. */
-Expr evalFile(EvalState & state, const Path & path);
 
-/* Evaluate an expression, and recursively evaluate list elements and
-   attributes.  If `canonicalise' is true, we remove things like
-   position information and make sure that attribute sets are in
-   sorded order. */
-Expr strictEvalExpr(EvalState & state, Expr e);
-
-/* Specific results. */
-string evalString(EvalState & state, Expr e, PathSet & context);
-string evalStringNoCtx(EvalState & state, Expr e);
-int evalInt(EvalState & state, Expr e);
-bool evalBool(EvalState & state, Expr e);
-ATermList evalList(EvalState & state, Expr e);
-
-/* Flatten nested lists into a single list (or expand a singleton into
-   a list). */
-ATermList flattenList(EvalState & state, Expr e);
-
-/* String coercion.  Converts strings, paths and derivations to a
-   string.  If `coerceMore' is set, also converts nulls, integers,
-   booleans and lists to a string. */
-string coerceToString(EvalState & state, Expr e, PathSet & context,
-    bool coerceMore = false, bool copyToStore = true);
-
-/* Path coercion.  Converts strings, paths and derivations to a path.
-   The result is guaranteed to be an canonicalised, absolute path.
-   Nothing is copied to the store. */
-Path coerceToPath(EvalState & state, Expr e, PathSet & context);
-
-/* Automatically call a function for which each argument has a default
-   value or has a binding in the `args' map.  Note: result is a call,
-   not a normal form; it should be evaluated by calling evalExpr(). */
-Expr autoCallFunction(Expr e, const ATermMap & args);
-
-/* Print statistics. */
-void printEvalStats(EvalState & state);
-
- 
 }
 
 

src/libexpr/expr-to-xml.cc

@@ -1,156 +0,0 @@
-#include "expr-to-xml.hh"
-#include "xml-writer.hh"
-#include "nixexpr-ast.hh"
-#include "aterm.hh"
-#include "util.hh"
-
-#include <cstdlib>
-
-
-namespace nix {
-
-    
-static XMLAttrs singletonAttrs(const string & name, const string & value)
-{
-    XMLAttrs attrs;
-    attrs[name] = value;
-    return attrs;
-}
-
-
-/* set<Expr> is safe because all the expressions are also reachable
-   from the stack, therefore can't be garbage-collected. */
-typedef set<Expr> ExprSet;
-
-
-static void printTermAsXML(Expr e, XMLWriter & doc, PathSet & context,
-    ExprSet & drvsSeen);
-
-
-static void showAttrs(const ATermMap & attrs, XMLWriter & doc,
-    PathSet & context, ExprSet & drvsSeen)
-{
-    StringSet names;
-    for (ATermMap::const_iterator i = attrs.begin(); i != attrs.end(); ++i)
-        names.insert(aterm2String(i->key));
-    for (StringSet::iterator i = names.begin(); i != names.end(); ++i) {
-        XMLOpenElement _(doc, "attr", singletonAttrs("name", *i));
-        printTermAsXML(attrs.get(toATerm(*i)), doc, context, drvsSeen);
-    }
-}
-
-
-static void printPatternAsXML(Pattern pat, XMLWriter & doc)
-{
-    ATerm name;
-    ATermList formals;
-    Pattern pat1, pat2;
-    ATermBool ellipsis;
-    if (matchVarPat(pat, name))
-        doc.writeEmptyElement("varpat", singletonAttrs("name", aterm2String(name)));
-    else if (matchAttrsPat(pat, formals, ellipsis)) {
-        XMLOpenElement _(doc, "attrspat");
-        for (ATermIterator i(formals); i; ++i) {
-            Expr name; ATerm dummy;
-            if (!matchFormal(*i, name, dummy)) abort();
-            doc.writeEmptyElement("attr", singletonAttrs("name", aterm2String(name)));
-        }
-        if (ellipsis == eTrue) doc.writeEmptyElement("ellipsis");
-    }
-    else if (matchAtPat(pat, pat1, pat2)) {
-        XMLOpenElement _(doc, "at");
-        printPatternAsXML(pat1, doc);
-        printPatternAsXML(pat2, doc);
-    }
-}
-
-
-static void printTermAsXML(Expr e, XMLWriter & doc, PathSet & context,
-    ExprSet & drvsSeen)
-{
-    XMLAttrs attrs;
-    string s;
-    ATerm s2;
-    int i;
-    ATermList as, es;
-    ATerm pat, body, pos;
-
-    checkInterrupt();
-
-    if (matchStr(e, s, context)) /* !!! show the context? */
-        doc.writeEmptyElement("string", singletonAttrs("value", s));
-
-    else if (matchPath(e, s2))
-        doc.writeEmptyElement("path", singletonAttrs("value", aterm2String(s2)));
-
-    else if (matchNull(e))
-        doc.writeEmptyElement("null");
-
-    else if (matchInt(e, i))
-        doc.writeEmptyElement("int", singletonAttrs("value", (format("%1%") % i).str()));
-
-    else if (e == eTrue)
-        doc.writeEmptyElement("bool", singletonAttrs("value", "true"));
-
-    else if (e == eFalse)
-        doc.writeEmptyElement("bool", singletonAttrs("value", "false"));
-
-    else if (matchAttrs(e, as)) {
-        ATermMap attrs;
-        queryAllAttrs(e, attrs);
-
-        Expr a = attrs.get(toATerm("type"));
-        if (a && matchStr(a, s, context) && s == "derivation") {
-
-            XMLAttrs xmlAttrs;
-            Path outPath, drvPath;
-            
-            a = attrs.get(toATerm("drvPath"));
-            if (matchStr(a, drvPath, context))
-                xmlAttrs["drvPath"] = drvPath;
-        
-            a = attrs.get(toATerm("outPath"));
-            if (matchStr(a, outPath, context))
-                xmlAttrs["outPath"] = outPath;
-        
-            XMLOpenElement _(doc, "derivation", xmlAttrs);
-
-            if (drvsSeen.find(e) == drvsSeen.end()) {
-                drvsSeen.insert(e);
-                showAttrs(attrs, doc, context, drvsSeen);
-            } else
-                doc.writeEmptyElement("repeated");
-        }
-
-        else {
-            XMLOpenElement _(doc, "attrs");
-            showAttrs(attrs, doc, context, drvsSeen);
-        }
-    }
-
-    else if (matchList(e, es)) {
-        XMLOpenElement _(doc, "list");
-        for (ATermIterator i(es); i; ++i)
-            printTermAsXML(*i, doc, context, drvsSeen);
-    }
-
-    else if (matchFunction(e, pat, body, pos)) {
-        XMLOpenElement _(doc, "function");
-        printPatternAsXML(pat, doc);
-    }
-
-    else
-        doc.writeEmptyElement("unevaluated");
-}
-
-
-void printTermAsXML(Expr e, std::ostream & out, PathSet & context)
-{
-    XMLWriter doc(true, out);
-    XMLOpenElement root(doc, "expr");
-    ExprSet drvsSeen;    
-    printTermAsXML(e, doc, context, drvsSeen);
-}
-
- 
-}

src/libexpr/expr-to-xml.hh

@@ -1,16 +0,0 @@
-#ifndef __EXPR_TO_XML_H
-#define __EXPR_TO_XML_H
-
-#include <string>
-#include <map>
-
-#include "nixexpr.hh"
-#include "aterm.hh"
-
-namespace nix {
-
-void printTermAsXML(Expr e, std::ostream & out, PathSet & context);
-    
-}
-
-#endif /* !__EXPR_TO_XML_H */

src/libexpr/get-drvs.cc

@@ -1,5 +1,4 @@
 #include "get-drvs.hh"
-#include "nixexpr-ast.hh"
 #include "util.hh"
 
 
@@ -8,17 +7,10 @@ namespace nix {
 
 string DrvInfo::queryDrvPath(EvalState & state) const
 {
-    if (drvPath == "") {
-        Expr a = attrs->get(toATerm("drvPath"));
-
-        /* Backwards compatibility hack with user environments made by
-           Nix <= 0.10: these contain illegal Path("") expressions. */
-        ATerm t;
-        if (a && matchPath(evalExpr(state, a), t))
-            return aterm2String(t);
-        
+    if (drvPath == "" && attrs) {
+        Bindings::iterator i = attrs->find(state.sDrvPath);
         PathSet context;
-        (string &) drvPath = a ? coerceToPath(state, a, context) : "";
+        (string &) drvPath = i != attrs->end() ? state.coerceToPath(*i->value, context) : "";
     }
     return drvPath;
 }
@@ -26,11 +18,10 @@ string DrvInfo::queryDrvPath(EvalState & state) const
 
 string DrvInfo::queryOutPath(EvalState & state) const
 {
-    if (outPath == "") {
-        Expr a = attrs->get(toATerm("outPath"));
-        if (!a) throw TypeError("output path missing");
+    if (outPath == "" && attrs) {
+        Bindings::iterator i = attrs->find(state.sOutPath);
         PathSet context;
-        (string &) outPath = coerceToPath(state, a, context);
+        (string &) outPath = i != attrs->end() ? state.coerceToPath(*i->value, context) : "";
     }
     return outPath;
 }
@@ -38,35 +29,30 @@ string DrvInfo::queryOutPath(EvalState & state) const
 
 MetaInfo DrvInfo::queryMetaInfo(EvalState & state) const
 {
-    MetaInfo meta;
+    if (metaInfoRead) return meta;
     
-    Expr a = attrs->get(toATerm("meta"));
-    if (!a) return meta; /* fine, empty meta information */
+    (bool &) metaInfoRead = true;
+    
+    Bindings::iterator a = attrs->find(state.sMeta);
+    if (a == attrs->end()) return meta; /* fine, empty meta information */
 
-    ATermMap attrs2;
-    queryAllAttrs(evalExpr(state, a), attrs2);
+    state.forceAttrs(*a->value);
 
-    for (ATermMap::const_iterator i = attrs2.begin(); i != attrs2.end(); ++i) {
-        Expr e = evalExpr(state, i->value);
-        string s;
-        PathSet context;
+    foreach (Bindings::iterator, i, *a->value->attrs) {
         MetaValue value;
-        int n;
-        ATermList es;
-        if (matchStr(e, s, context)) {
+        state.forceValue(*i->value);
+        if (i->value->type == tString) {
             value.type = MetaValue::tpString;
-            value.stringValue = s;
-            meta[aterm2String(i->key)] = value;
-        } else if (matchInt(e, n)) {
+            value.stringValue = i->value->string.s;
+        } else if (i->value->type == tInt) {
             value.type = MetaValue::tpInt;
-            value.intValue = n;
-            meta[aterm2String(i->key)] = value;
-        } else if (matchList(e, es)) {
+            value.intValue = i->value->integer;
+        } else if (i->value->type == tList) {
             value.type = MetaValue::tpStrings;
-            for (ATermIterator j(es); j; ++j)
-                value.stringValues.push_back(evalStringNoCtx(state, *j));
-            meta[aterm2String(i->key)] = value;
-        }
+            for (unsigned int j = 0; j < i->value->list.length; ++j)
+                value.stringValues.push_back(state.forceStringNoCtx(*i->value->list.elems[j]));
+        } else continue;
+        ((MetaInfo &) meta)[i->name] = value;
     }
 
     return meta;
@@ -82,73 +68,46 @@ MetaValue DrvInfo::queryMetaInfo(EvalState & state, const string & name) const
 
 void DrvInfo::setMetaInfo(const MetaInfo & meta)
 {
-    ATermMap metaAttrs;
-    foreach (MetaInfo::const_iterator, i, meta) {
-        Expr e;
-        switch (i->second.type) {
-            case MetaValue::tpInt: e = makeInt(i->second.intValue); break;
-            case MetaValue::tpString: e = makeStr(i->second.stringValue); break;
-            case MetaValue::tpStrings: {
-                ATermList es = ATempty;
-                foreach (Strings::const_iterator, j, i->second.stringValues)
-                    es = ATinsert(es, makeStr(*j));
-                e = makeList(ATreverse(es));
-                break;
-            }
-            default: abort();
-        }
-        metaAttrs.set(toATerm(i->first), makeAttrRHS(e, makeNoPos()));
-    }
-    attrs->set(toATerm("meta"), makeAttrs(metaAttrs));
+    metaInfoRead = true;
+    this->meta = meta;
 }
 
 
-/* Cache for already evaluated derivations.  Usually putting ATerms in
-   a STL container is unsafe (they're not scanning for GC roots), but
-   here it doesn't matter; everything in this set is reachable from
-   the stack as well. */
-typedef set<Expr> Exprs;
+/* Cache for already considered attrsets. */
+typedef set<Bindings *> Done;
 
 
-/* Evaluate expression `e'.  If it evaluates to an attribute set of
-   type `derivation', then put information about it in `drvs' (unless
-   it's already in `doneExprs').  The result boolean indicates whether
-   it makes sense for the caller to recursively search for derivations
-   in `e'. */
-static bool getDerivation(EvalState & state, Expr e,
-    const string & attrPath, DrvInfos & drvs, Exprs & doneExprs)
+/* Evaluate value `v'.  If it evaluates to an attribute set of type
+   `derivation', then put information about it in `drvs' (unless it's
+   already in `doneExprs').  The result boolean indicates whether it
+   makes sense for the caller to recursively search for derivations in
+   `v'. */
+static bool getDerivation(EvalState & state, Value & v,
+    const string & attrPath, DrvInfos & drvs, Done & done)
 {
     try {
-        
-        ATermList es;
-        e = evalExpr(state, e);
-        if (!matchAttrs(e, es)) return true;
-
-        boost::shared_ptr<ATermMap> attrs(new ATermMap());
-        queryAllAttrs(e, *attrs, false);
-        
-        Expr a = attrs->get(toATerm("type"));
-        if (!a || evalStringNoCtx(state, a) != "derivation") return true;
+        state.forceValue(v);
+        if (!state.isDerivation(v)) return true;
 
         /* Remove spurious duplicates (e.g., an attribute set like
            `rec { x = derivation {...}; y = x;}'. */
-        if (doneExprs.find(e) != doneExprs.end()) return false;
-        doneExprs.insert(e);
+        if (done.find(v.attrs) != done.end()) return false;
+        done.insert(v.attrs);
 
         DrvInfo drv;
     
-        a = attrs->get(toATerm("name"));
+        Bindings::iterator i = v.attrs->find(state.sName);
         /* !!! We really would like to have a decent back trace here. */
-        if (!a) throw TypeError("derivation name missing");
-        drv.name = evalStringNoCtx(state, a);
+        if (i == v.attrs->end()) throw TypeError("derivation name missing");
+        drv.name = state.forceStringNoCtx(*i->value);
 
-        a = attrs->get(toATerm("system"));
-        if (!a)
+        Bindings::iterator i2 = v.attrs->find(state.sSystem);
+        if (i2 == v.attrs->end())
             drv.system = "unknown";
         else
-            drv.system = evalStringNoCtx(state, a);
+            drv.system = state.forceStringNoCtx(*i2->value);
 
-        drv.attrs = attrs;
+        drv.attrs = v.attrs;
 
         drv.attrPath = attrPath;
 
@@ -161,11 +120,11 @@ static bool getDerivation(EvalState & state, Expr e,
 }
 
 
-bool getDerivation(EvalState & state, Expr e, DrvInfo & drv)
+bool getDerivation(EvalState & state, Value & v, DrvInfo & drv)
 {
-    Exprs doneExprs;
+    Done done;
     DrvInfos drvs;
-    getDerivation(state, e, "", drvs, doneExprs);
+    getDerivation(state, v, "", drvs, done);
     if (drvs.size() != 1) return false;
     drv = drvs.front();
     return true;
@@ -178,83 +137,73 @@ static string addToPath(const string & s1, const string & s2)
 }
 
 
-static void getDerivations(EvalState & state, Expr e,
-    const string & pathPrefix, const ATermMap & autoArgs,
-    DrvInfos & drvs, Exprs & doneExprs)
+static void getDerivations(EvalState & state, Value & vIn,
+    const string & pathPrefix, Bindings & autoArgs,
+    DrvInfos & drvs, Done & done)
 {
-    e = evalExpr(state, autoCallFunction(evalExpr(state, e), autoArgs));
-
+    Value v;
+    state.autoCallFunction(autoArgs, vIn, v);
+    
     /* Process the expression. */
-    ATermList es;
     DrvInfo drv;
 
-    if (!getDerivation(state, e, pathPrefix, drvs, doneExprs))
-        return;
+    if (!getDerivation(state, v, pathPrefix, drvs, done)) ;
 
-    if (matchAttrs(e, es)) {
-        ATermMap drvMap(ATgetLength(es));
-        queryAllAttrs(e, drvMap);
+    else if (v.type == tAttrs) {
 
         /* !!! undocumented hackery to support combining channels in
            nix-env.cc. */
-        bool combineChannels = drvMap.get(toATerm("_combineChannels"));
+        bool combineChannels = v.attrs->find(state.symbols.create("_combineChannels")) != v.attrs->end();
 
         /* Consider the attributes in sorted order to get more
            deterministic behaviour in nix-env operations (e.g. when
            there are names clashes between derivations, the derivation
            bound to the attribute with the "lower" name should take
            precedence). */
-        typedef std::map<string, Expr> AttrsSorted;
-        AttrsSorted attrsSorted;
-        foreach (ATermMap::const_iterator, i, drvMap)
-            attrsSorted[aterm2String(i->key)] = i->value;
+        typedef std::map<string, Symbol> SortedSymbols;
+        SortedSymbols attrs;
+        foreach (Bindings::iterator, i, *v.attrs)
+            attrs.insert(std::pair<string, Symbol>(i->name, i->name));
 
-        foreach (AttrsSorted::iterator, i, attrsSorted) {
+        foreach (SortedSymbols::iterator, i, attrs) {
             startNest(nest, lvlDebug, format("evaluating attribute `%1%'") % i->first);
             string pathPrefix2 = addToPath(pathPrefix, i->first);
+            Value & v2(*v.attrs->find(i->second)->value);
             if (combineChannels)
-                getDerivations(state, i->second, pathPrefix2, autoArgs, drvs, doneExprs);
-            else if (getDerivation(state, i->second, pathPrefix2, drvs, doneExprs)) {
+                getDerivations(state, v2, pathPrefix2, autoArgs, drvs, done);
+            else if (getDerivation(state, v2, pathPrefix2, drvs, done)) {
                 /* If the value of this attribute is itself an
                    attribute set, should we recurse into it?  => Only
                    if it has a `recurseForDerivations = true'
                    attribute. */
-                ATermList es;
-                Expr e = evalExpr(state, i->second), e2;
-                if (matchAttrs(e, es)) {
-                    ATermMap attrs(ATgetLength(es));
-                    queryAllAttrs(e, attrs, false);
-                    if (((e2 = attrs.get(toATerm("recurseForDerivations")))
-                            && evalBool(state, e2)))
-                        getDerivations(state, e, pathPrefix2, autoArgs, drvs, doneExprs);
+                if (v2.type == tAttrs) {
+                    Bindings::iterator j = v2.attrs->find(state.symbols.create("recurseForDerivations"));
+                    if (j != v2.attrs->end() && state.forceBool(*j->value))
+                        getDerivations(state, v2, pathPrefix2, autoArgs, drvs, done);
                 }
             }
         }
-        
-        return;
     }
 
-    if (matchList(e, es)) {
-        int n = 0;
-        for (ATermIterator i(es); i; ++i, ++n) {
+    else if (v.type == tList) {
+        for (unsigned int n = 0; n < v.list.length; ++n) {
             startNest(nest, lvlDebug,
                 format("evaluating list element"));
             string pathPrefix2 = addToPath(pathPrefix, (format("%1%") % n).str());
-            if (getDerivation(state, *i, pathPrefix2, drvs, doneExprs))
-                getDerivations(state, *i, pathPrefix2, autoArgs, drvs, doneExprs);
+            if (getDerivation(state, *v.list.elems[n], pathPrefix2, drvs, done))
+                getDerivations(state, *v.list.elems[n], pathPrefix2, autoArgs, drvs, done);
         }
-        return;
     }
 
-    throw TypeError("expression does not evaluate to a derivation (or a set or list of those)");
+    else throw TypeError("expression does not evaluate to a derivation (or a set or list of those)");
 }
 
 
-void getDerivations(EvalState & state, Expr e, const string & pathPrefix,
-    const ATermMap & autoArgs, DrvInfos & drvs)
+void getDerivations(EvalState & state, Value & v, const string & pathPrefix,
+    Bindings & autoArgs, DrvInfos & drvs)
 {
-    Exprs doneExprs;
-    getDerivations(state, e, pathPrefix, autoArgs, drvs, doneExprs);
+    Done done;
+    getDerivations(state, v, pathPrefix, autoArgs, drvs, done);
 }
 
  

src/libexpr/get-drvs.hh

@@ -1,13 +1,13 @@
 #ifndef __GET_DRVS_H
 #define __GET_DRVS_H
 
+#include "eval.hh"
+
 #include <string>
 #include <map>
 
 #include <boost/shared_ptr.hpp>
 
-#include "eval.hh"
-
 
 namespace nix {
 
@@ -29,16 +29,19 @@ struct DrvInfo
 private:
     string drvPath;
     string outPath;
+
+    bool metaInfoRead;
+    MetaInfo meta;
     
 public:
     string name;
     string attrPath; /* path towards the derivation */
     string system;
 
-    /* !!! these should really be hidden, and setMetaInfo() should
-       make a copy since the ATermMap can be shared between multiple
-       DrvInfos. */
-    boost::shared_ptr<ATermMap> attrs;
+    /* !!! make this private */
+    Bindings * attrs;
+
+    DrvInfo() : metaInfoRead(false), attrs(0) { };
 
     string queryDrvPath(EvalState & state) const;
     string queryOutPath(EvalState & state) const;
@@ -59,16 +62,19 @@ public:
 };
 
 
+#if HAVE_BOEHMGC
+typedef list<DrvInfo, traceable_allocator<DrvInfo> > DrvInfos;
+#else
 typedef list<DrvInfo> DrvInfos;
+#endif
 
 
-/* Evaluate expression `e'.  If it evaluates to a derivation, store
-   information about the derivation in `drv' and return true.
-   Otherwise, return false. */
-bool getDerivation(EvalState & state, Expr e, DrvInfo & drv);
+/* If value `v' denotes a derivation, store information about the
+   derivation in `drv' and return true.  Otherwise, return false. */
+bool getDerivation(EvalState & state, Value & v, DrvInfo & drv);
 
-void getDerivations(EvalState & state, Expr e, const string & pathPrefix,
-    const ATermMap & autoArgs, DrvInfos & drvs);
+void getDerivations(EvalState & state, Value & v, const string & pathPrefix,
+    Bindings & autoArgs, DrvInfos & drvs);
 
  
 }

src/libexpr/lexer.l

@@ -8,9 +8,7 @@
 
 
 %{
-#include "aterm.hh"
 #include "nixexpr.hh"
-#include "nixexpr-ast.hh"
 #define BISON_HEADER_HACK
 #include "parser-tab.hh"
 
@@ -21,13 +19,16 @@ namespace nix {
     
 static void initLoc(YYLTYPE * loc)
 {
-    loc->first_line = 1;
-    loc->first_column = 1;
+    loc->first_line = loc->last_line = 1;
+    loc->first_column = loc->last_column = 1;
 }
 
     
 static void adjustLoc(YYLTYPE * loc, const char * s, size_t len)
 {
+    loc->first_line = loc->last_line;
+    loc->first_column = loc->last_column;
+
     while (len--) {
        switch (*s++) {
        case '\r':
@@ -35,17 +36,17 @@ static void adjustLoc(YYLTYPE * loc, const char * s, size_t len)
                s++;
            /* fall through */
        case '\n': 
-           ++loc->first_line;
-           loc->first_column = 1;
+           ++loc->last_line;
+           loc->last_column = 1;
            break;
        default:
-           ++loc->first_column;
+           ++loc->last_column;
        }
     }
 }
 
 
-static Expr unescapeStr(const char * s)
+static Expr * unescapeStr(SymbolTable & symbols, const char * s)
 {
     string t;
     char c;
@@ -65,7 +66,7 @@ static Expr unescapeStr(const char * s)
         }
         else t += c;
     }
-    return makeStr(toATerm(t), ATempty);
+    return new ExprString(symbols.create(t));
 }
 
  
@@ -105,19 +106,20 @@ inherit     { return INHERIT; }
 \/\/        { return UPDATE; }
 \+\+        { return CONCAT; }
 
-{ID}        { yylval->t = toATerm(yytext); return ID; /* !!! alloc */ }
+{ID}        { yylval->id = strdup(yytext); return ID; }
 {INT}       { int n = atoi(yytext); /* !!! overflow */
-              yylval->t = ATmake("<int>", n);
+              yylval->n = n;
               return INT;
             }
 
 \"          { BEGIN(STRING); return '"'; }
 <STRING>([^\$\"\\]|\$[^\{\"]|\\.)+ {
-/* !!! Not quite right: we want a follow restriction on "$", it
-   shouldn't be followed by a "{".  Right now "$\"" will be consumed
-   as part of a string, rather than a "$" followed by the string
-   terminator.  Disallow "$\"" for now. */
-              yylval->t = unescapeStr(yytext); /* !!! alloc */ 
+              /* !!! Not quite right: we want a follow restriction on
+                 "$", it shouldn't be followed by a "{".  Right now
+                 "$\"" will be consumed as part of a string, rather
+                 than a "$" followed by the string terminator.
+                 Disallow "$\"" for now. */
+              yylval->e = unescapeStr(data->symbols, yytext);
               return STR;
             }
 <STRING>\$\{  { BEGIN(INITIAL); return DOLLAR_CURLY; }
@@ -126,31 +128,31 @@ inherit     { return INHERIT; }
 
 \'\'(\ *\n)?     { BEGIN(IND_STRING); return IND_STRING_OPEN; }
 <IND_STRING>([^\$\']|\$[^\{\']|\'[^\'\$])+ {
-                   yylval->t = makeIndStr(toATerm(yytext));
+                   yylval->e = new ExprIndStr(yytext);
                    return IND_STR;
                  }
 <IND_STRING>\'\'\$ {
-                   yylval->t = makeIndStr(toATerm("$"));
+                   yylval->e = new ExprIndStr("$");
                    return IND_STR;
                  }
 <IND_STRING>\'\'\' {
-                   yylval->t = makeIndStr(toATerm("''"));
+                   yylval->e = new ExprIndStr("''");
                    return IND_STR;
                  }
 <IND_STRING>\'\'\\. {
-                   yylval->t = unescapeStr(yytext + 2);
+                   yylval->e = unescapeStr(data->symbols, yytext + 2);
                    return IND_STR;
                  }
 <IND_STRING>\$\{ { BEGIN(INITIAL); return DOLLAR_CURLY; }
 <IND_STRING>\'\' { BEGIN(INITIAL); return IND_STRING_CLOSE; }
 <IND_STRING>\'   {
-                   yylval->t = makeIndStr(toATerm("'"));
+                   yylval->e = new ExprIndStr("'");
                    return IND_STR;
                  }
 <IND_STRING>.    return yytext[0]; /* just in case: shouldn't be reached */
 
-{PATH}      { yylval->t = toATerm(yytext); return PATH; /* !!! alloc */ }
-{URI}       { yylval->t = toATerm(yytext); return URI; /* !!! alloc */ }
+{PATH}      { yylval->path = strdup(yytext); return PATH; }
+{URI}       { yylval->uri = strdup(yytext); return URI; }
 
 [ \t\r\n]+    /* eat up whitespace */
 \#[^\r\n]*    /* single-line comments */

src/libexpr/nixexpr-ast.def

@@ -1,97 +0,0 @@
-init initNixExprHelpers
-
-Pos | string int int | Pos |
-NoPos | | Pos |
-
-Function | Pattern Expr Pos | Expr |
-Assert | Expr Expr Pos | Expr |
-With | Expr Expr Pos | Expr |
-If | Expr Expr Expr | Expr |
-OpNot | Expr | Expr |
-OpEq | Expr Expr | Expr |
-OpNEq | Expr Expr | Expr |
-OpAnd | Expr Expr | Expr |
-OpOr | Expr Expr | Expr |
-OpImpl | Expr Expr | Expr |
-OpUpdate | Expr Expr | Expr |
-SubPath | Expr Expr | Expr |
-OpHasAttr | Expr string | Expr |
-OpPlus | Expr Expr | Expr |
-OpConcat | Expr Expr | Expr |
-ConcatStrings | ATermList | Expr |
-Call | Expr Expr | Expr |
-Select | Expr string | Expr |
-Var | string | Expr |
-Int | int | Expr |
-
-# Strings in the evaluator carry a so-called `context' (the ATermList)
-# which is a list of strings representing store paths.  This is to
-# allow users to write things like
-#
-#   "--with-freetype2-library=" + freetype + "/lib"
-#
-# where `freetype' is a derivation (or a source to be copied to the
-# store).  If we just concatenated the strings without keeping track
-# of the referenced store paths, then if the string is used as a
-# derivation attribute, the derivation will not have the correct
-# dependencies in its inputDrvs and inputSrcs.
-#
-# The semantics of the context is as follows: when a string with
-# context C is used as a derivation attribute, then the derivations in
-# C will be added to the inputDrvs of the derivation, and the other
-# store paths in C will be added to the inputSrcs of the derivations.
-#
-# For canonicity, the store paths should be in sorted order.
-Str | string ATermList | Expr |
-Str | string | Expr | ObsoleteStr
-
-# Internal to the parser, doesn't occur in ASTs.
-IndStr | string | Expr |
-
-# A path is a reference to a file system object that is to be copied
-# to the Nix store when used as a derivation attribute.  When it is
-# concatenated to a string (i.e., `str + path'), it is also copied and
-# the resulting store path is concatenated to the string (with the
-# store path in the context).  If a string or path is concatenated to
-# a path (i.e., `path + str' or `path + path'), the result is a new
-# path (if the right-hand side is a string, the context must be
-# empty).
-Path | string | Expr |
-
-List | ATermList | Expr |
-BlackHole | | Expr |
-Undefined | | Expr |
-Removed | | Expr |
-PrimOp | int ATermBlob ATermList | Expr |
-Attrs | ATermList | Expr |
-Closed | Expr | Expr |
-Rec | ATermList ATermList | Expr |
-Bool | ATermBool | Expr |
-Null | | Expr |
-
-Bind | string Expr Pos | ATerm |
-BindAttrPath | ATermList Expr Pos | ATerm | # desugared during parsing
-Bind | string Expr | ATerm | ObsoleteBind
-Inherit | Expr ATermList Pos | ATerm |
-
-Scope | | Expr |
-
-VarPat | string | Pattern |
-AttrsPat | ATermList ATermBool | Pattern | # bool = `...'
-AtPat | Pattern Pattern | Pattern |
-
-Formal | string DefaultValue | ATerm |
-
-DefaultValue | Expr | DefaultValue |
-NoDefaultValue | | DefaultValue |
-
-True | | ATermBool |
-False | | ATermBool |
-
-PrimOpDef | int ATermBlob | ATerm |
-
-AttrRHS | Expr Pos | ATerm |
-
-eTrue = makeBool(makeTrue())
-eFalse = makeBool(makeFalse())
-sOverrides = toATerm("__overrides")

src/libexpr/nixexpr.cc

@@ -1,407 +1,314 @@
 #include "nixexpr.hh"
 #include "derivations.hh"
 #include "util.hh"
-#include "aterm.hh"
-
-#include "nixexpr-ast.hh"
-#include "nixexpr-ast.cc"
 
 #include <cstdlib>
 
 
 namespace nix {
-    
 
-string showPos(ATerm pos)
+
+/* Displaying abstract syntax trees. */
+
+std::ostream & operator << (std::ostream & str, Expr & e)
 {
-    ATerm path;
-    int line, column;
-    if (matchNoPos(pos)) return "undefined position";
-    if (!matchPos(pos, path, line, column))
-        throw badTerm("position expected", pos);
-    return (format("`%1%:%2%:%3%'") % aterm2String(path) % line % column).str();
+    e.show(str);
+    return str;
 }
-    
 
-ATerm bottomupRewrite(TermFun & f, ATerm e)
+void Expr::show(std::ostream & str)
 {
-    checkInterrupt();
+    abort();
+}
 
-    if (ATgetType(e) == AT_APPL) {
-        AFun fun = ATgetAFun(e);
-        int arity = ATgetArity(fun);
-        ATerm args[arity];
+void ExprInt::show(std::ostream & str)
+{
+    str << n;
+}
 
-        for (int i = 0; i < arity; ++i)
-            args[i] = bottomupRewrite(f, ATgetArgument(e, i));
+void ExprString::show(std::ostream & str)
+{
+    str << "\"" << s << "\""; // !!! escaping
+}
+
+void ExprPath::show(std::ostream & str)
+{
+    str << s;
+}
+
+void ExprVar::show(std::ostream & str)
+{
+    str << info.name;
+}
+
+void ExprSelect::show(std::ostream & str)
+{
+    str << "(" << *e << ")." << name;
+}
+
+void ExprOpHasAttr::show(std::ostream & str)
+{
+    str << "(" << *e << ") ? " << name;
+}
+
+void ExprAttrs::show(std::ostream & str)
+{
+    if (recursive) str << "rec ";
+    str << "{ ";
+    foreach (AttrDefs::iterator, i, attrs)
+        if (i->second.inherited)
+            str << "inherit " << i->first << " " << "; ";
+        else
+            str << i->first << " = " << *i->second.e << "; ";
+    str << "}";
+}
+
+void ExprList::show(std::ostream & str)
+{
+    str << "[ ";
+    foreach (vector<Expr *>::iterator, i, elems)
+        str << "(" << **i << ") ";
+    str << "]";
+}
+
+void ExprLambda::show(std::ostream & str)
+{
+    str << "(";
+    if (matchAttrs) {
+        str << "{ ";
+        bool first = true;
+        foreach (Formals::Formals_::iterator, i, formals->formals) {
+            if (first) first = false; else str << ", ";
+            str << i->name;
+            if (i->def) str << " ? " << *i->def;
+        }
+        str << " }";
+        if (!arg.empty()) str << " @ ";
+    }
+    if (!arg.empty()) str << arg;
+    str << ": " << *body << ")";
+}
+
+void ExprLet::show(std::ostream & str)
+{
+    str << "let ";
+    foreach (ExprAttrs::AttrDefs::iterator, i, attrs->attrs)
+        if (i->second.inherited)
+            str << "inherit " << i->first << "; ";
+        else
+            str << i->first << " = " << *i->second.e << "; ";
+    str << "in " << *body;
+}
+
+void ExprWith::show(std::ostream & str)
+{
+    str << "with " << *attrs << "; " << *body;
+}
+
+void ExprIf::show(std::ostream & str)
+{
+    str << "if " << *cond << " then " << *then << " else " << *else_;
+}
+
+void ExprAssert::show(std::ostream & str)
+{
+    str << "assert " << *cond << "; " << *body;
+}
+
+void ExprOpNot::show(std::ostream & str)
+{
+    str << "! " << *e;
+}
+
+void ExprConcatStrings::show(std::ostream & str)
+{
+    bool first = true;
+    foreach (vector<Expr *>::iterator, i, *es) {
+        if (first) first = false; else str << " + ";
+        str << **i;
+    }
+}
+
+
+std::ostream & operator << (std::ostream & str, const Pos & pos)
+{
+    if (!pos.line)
+        str << "undefined position";
+    else
+        str << (format("`%1%:%2%:%3%'") % pos.file % pos.line % pos.column).str();
+    return str;
+}
+
+
+Pos noPos;
+
+
+/* Computing levels/displacements for variables. */
+
+void Expr::bindVars(const StaticEnv & env)
+{
+    abort();
+}
+
+void ExprInt::bindVars(const StaticEnv & env)
+{
+}
+
+void ExprString::bindVars(const StaticEnv & env)
+{
+}
+
+void ExprPath::bindVars(const StaticEnv & env)
+{
+}
+
+void VarRef::bind(const StaticEnv & env)
+{
+    /* Check whether the variable appears in the environment.  If so,
+       set its level and displacement. */
+    const StaticEnv * curEnv;
+    unsigned int level;
+    int withLevel = -1;
+    for (curEnv = &env, level = 0; curEnv; curEnv = curEnv->up, level++) {
+        if (curEnv->isWith) {
+            if (withLevel == -1) withLevel = level;
+        } else {
+            StaticEnv::Vars::const_iterator i = curEnv->vars.find(name);
+            if (i != curEnv->vars.end()) {
+                fromWith = false;
+                this->level = level;
+                displ = i->second;
+                return;
+            }
+        }
+    }
+
+    /* Otherwise, the variable must be obtained from the nearest
+       enclosing `with'.  If there is no `with', then we can issue an
+       "undefined variable" error now. */
+    if (withLevel == -1) throw EvalError(format("undefined variable `%1%'") % name);
+
+    fromWith = true;
+    this->level = withLevel;
+}
+
+void ExprVar::bindVars(const StaticEnv & env)
+{
+    info.bind(env);
+}
+
+void ExprSelect::bindVars(const StaticEnv & env)
+{
+    e->bindVars(env);
+}
+
+void ExprOpHasAttr::bindVars(const StaticEnv & env)
+{
+    e->bindVars(env);
+}
+
+void ExprAttrs::bindVars(const StaticEnv & env)
+{
+    if (recursive) {
+        StaticEnv newEnv(false, &env);
+    
+        unsigned int displ = 0;
+        foreach (AttrDefs::iterator, i, attrs)
+            newEnv.vars[i->first] = i->second.displ = displ++;
         
-        e = (ATerm) ATmakeApplArray(fun, args);
+        foreach (AttrDefs::iterator, i, attrs)
+            if (i->second.inherited) i->second.var.bind(env);
+            else i->second.e->bindVars(newEnv);
     }
 
-    else if (ATgetType(e) == AT_LIST) {
-        ATermList in = (ATermList) e;
-        ATermList out = ATempty;
-
-        for (ATermIterator i(in); i; ++i)
-            out = ATinsert(out, bottomupRewrite(f, *i));
-
-        e = (ATerm) ATreverse(out);
-    }
-
-    return f(e);
+    else
+        foreach (AttrDefs::iterator, i, attrs)
+            if (i->second.inherited) i->second.var.bind(env);
+            else i->second.e->bindVars(env);
 }
 
-
-void queryAllAttrs(Expr e, ATermMap & attrs, bool withPos)
+void ExprList::bindVars(const StaticEnv & env)
 {
-    ATermList bnds;
-    if (!matchAttrs(e, bnds))
-        throw TypeError(format("value is %1% while an attribute set was expected") % showType(e));
-
-    for (ATermIterator i(bnds); i; ++i) {
-        ATerm name;
-        Expr e;
-        ATerm pos;
-        if (!matchBind(*i, name, e, pos)) abort(); /* can't happen */
-        attrs.set(name, withPos ? makeAttrRHS(e, pos) : e);
-    }
+    foreach (vector<Expr *>::iterator, i, elems)
+        (*i)->bindVars(env);
 }
 
-
-Expr queryAttr(Expr e, const string & name)
+void ExprLambda::bindVars(const StaticEnv & env)
 {
-    ATerm dummy;
-    return queryAttr(e, name, dummy);
-}
-
-
-Expr queryAttr(Expr e, const string & name, ATerm & pos)
-{
-    ATermList bnds;
-    if (!matchAttrs(e, bnds))
-        throw TypeError(format("value is %1% while an attribute set was expected") % showType(e));
-
-    for (ATermIterator i(bnds); i; ++i) {
-        ATerm name2, pos2;
-        Expr e;
-        if (!matchBind(*i, name2, e, pos2))
-            abort(); /* can't happen */
-        if (aterm2String(name2) == name) {
-            pos = pos2;
-            return e;
-        }
-    }
-
-    return 0;
-}
-
-
-Expr makeAttrs(const ATermMap & attrs)
-{
-    ATermList bnds = ATempty;
-    for (ATermMap::const_iterator i = attrs.begin(); i != attrs.end(); ++i) {
-        Expr e;
-        ATerm pos;
-        if (!matchAttrRHS(i->value, e, pos))
-            abort(); /* can't happen */
-        bnds = ATinsert(bnds, makeBind(i->key, e, pos));
-    }
-    return makeAttrs(bnds);
-}
-
-
-static void varsBoundByPattern(ATermMap & map, Pattern pat)
-{
-    ATerm name;
-    ATermList formals;
-    Pattern pat1, pat2;
-    ATermBool ellipsis;
-    /* Use makeRemoved() so that it can be used directly in
-       substitute(). */
-    if (matchVarPat(pat, name))
-        map.set(name, makeRemoved());
-    else if (matchAttrsPat(pat, formals, ellipsis)) { 
-        for (ATermIterator i(formals); i; ++i) {
-            ATerm d1;
-            if (!matchFormal(*i, name, d1)) abort();
-            map.set(name, makeRemoved());
-        }
-    }
-    else if (matchAtPat(pat, pat1, pat2)) {
-        varsBoundByPattern(map, pat1);
-        varsBoundByPattern(map, pat2);
-    }
-    else abort();
-}
-
-
-Expr substitute(const Substitution & subs, Expr e)
-{
-    checkInterrupt();
-
-    //if (subs.size() == 0) return e;
-
-    ATerm name, pos, e2;
-
-    /* As an optimisation, don't substitute in subterms known to be
-       closed. */
-    if (matchClosed(e, e2)) return e;
-
-    if (matchVar(e, name)) {
-        Expr sub = subs.lookup(name);
-        if (sub == makeRemoved()) sub = 0;
-        Expr wrapped;
-        /* Add a "closed" wrapper around terms that aren't already
-           closed.  The check is necessary to prevent repeated
-           wrapping, e.g., closed(closed(closed(...))), which kills
-           caching. */
-        return sub ? (matchClosed(sub, wrapped) ? sub : makeClosed(sub)) : e;
-    }
-
-    /* In case of a function, filter out all variables bound by this
-       function. */
-    Pattern pat;
-    ATerm body;
-    if (matchFunction(e, pat, body, pos)) {
-        ATermMap map(16);
-        varsBoundByPattern(map, pat);
-        Substitution subs2(&subs, &map);
-        return makeFunction(
-            (Pattern) substitute(subs2, (Expr) pat),
-            substitute(subs2, body), pos);
-    }
-
-    /* Idem for a mutually recursive attribute set. */
-    ATermList rbnds, nrbnds;
-    if (matchRec(e, rbnds, nrbnds)) {
-        ATermMap map(ATgetLength(rbnds) + ATgetLength(nrbnds));
-        for (ATermIterator i(rbnds); i; ++i)
-            if (matchBind(*i, name, e2, pos)) map.set(name, makeRemoved());
-            else abort(); /* can't happen */
-        for (ATermIterator i(nrbnds); i; ++i)
-            if (matchBind(*i, name, e2, pos)) map.set(name, makeRemoved());
-            else abort(); /* can't happen */
-        return makeRec(
-            (ATermList) substitute(Substitution(&subs, &map), (ATerm) rbnds),
-            (ATermList) substitute(subs, (ATerm) nrbnds));
-    }
-
-    if (ATgetType(e) == AT_APPL) {
-        AFun fun = ATgetAFun(e);
-        int arity = ATgetArity(fun);
-        ATerm args[arity];
-        bool changed = false;
-
-        for (int i = 0; i < arity; ++i) {
-            ATerm arg = ATgetArgument(e, i);
-            args[i] = substitute(subs, arg);
-            if (args[i] != arg) changed = true;
-        }
-        
-        return changed ? (ATerm) ATmakeApplArray(fun, args) : e;
-    }
-
-    if (ATgetType(e) == AT_LIST) {
-        unsigned int len = ATgetLength((ATermList) e);
-        ATerm es[len];
-        ATermIterator i((ATermList) e);
-        for (unsigned int j = 0; i; ++i, ++j)
-            es[j] = substitute(subs, *i);
-        ATermList out = ATempty;
-        for (unsigned int j = len; j; --j)
-            out = ATinsert(out, es[j - 1]);
-        return (ATerm) out;
-    }
-
-    return e;
-}
-
-
-/* We use memoisation to prevent exponential complexity on heavily
-   shared ATerms (remember, an ATerm is a graph, not a tree!).  Note
-   that using an STL set is fine here wrt to ATerm garbage collection
-   since all the ATerms in the set are already reachable from
-   somewhere else. */
-static void checkVarDefs2(set<Expr> & done, const ATermMap & defs, Expr e)
-{
-    if (done.find(e) != done.end()) return;
-    done.insert(e);
+    StaticEnv newEnv(false, &env);
     
-    ATerm name, pos, value;
-    ATerm with, body;
-    ATermList rbnds, nrbnds;
-    Pattern pat;
-
-    /* Closed terms don't have free variables, so we don't have to
-       check by definition. */
-    if (matchClosed(e, value)) return;
+    unsigned int displ = 0;
     
-    else if (matchVar(e, name)) {
-        if (!defs.get(name))
-            throw EvalError(format("undefined variable `%1%'")
-                % aterm2String(name));
+    if (!arg.empty()) newEnv.vars[arg] = displ++;
+
+    if (matchAttrs) {
+        foreach (Formals::Formals_::iterator, i, formals->formals)
+            newEnv.vars[i->name] = displ++;
+
+        foreach (Formals::Formals_::iterator, i, formals->formals)
+            if (i->def) i->def->bindVars(newEnv);
     }
 
-    else if (matchFunction(e, pat, body, pos)) {
-        ATermMap defs2(defs);
-        varsBoundByPattern(defs2, pat);
-        set<Expr> done2;
-        checkVarDefs2(done2, defs2, pat);
-        checkVarDefs2(done2, defs2, body);
-    }
-        
-    else if (matchRec(e, rbnds, nrbnds)) {
-        checkVarDefs2(done, defs, (ATerm) nrbnds);
-        ATermMap defs2(defs);
-        for (ATermIterator i(rbnds); i; ++i) {
-            if (!matchBind(*i, name, value, pos)) abort(); /* can't happen */
-            defs2.set(name, (ATerm) ATempty);
-        }
-        for (ATermIterator i(nrbnds); i; ++i) {
-            if (!matchBind(*i, name, value, pos)) abort(); /* can't happen */
-            defs2.set(name, (ATerm) ATempty);
-        }
-        set<Expr> done2;
-        checkVarDefs2(done2, defs2, (ATerm) rbnds);
-    }
+    body->bindVars(newEnv);
+}
 
-    else if (matchWith(e, with, body, pos)) {
-        /* We can't check the body without evaluating the definitions
-           (which is an arbitrary expression), so we don't do that
-           here but only when actually evaluating the `with'. */
-        checkVarDefs2(done, defs, with);
-    }
+void ExprLet::bindVars(const StaticEnv & env)
+{
+    StaticEnv newEnv(false, &env);
     
-    else if (ATgetType(e) == AT_APPL) {
-        int arity = ATgetArity(ATgetAFun(e));
-        for (int i = 0; i < arity; ++i)
-            checkVarDefs2(done, defs, ATgetArgument(e, i));
-    }
-
-    else if (ATgetType(e) == AT_LIST)
-        for (ATermIterator i((ATermList) e); i; ++i)
-            checkVarDefs2(done, defs, *i);
+    unsigned int displ = 0;
+    foreach (ExprAttrs::AttrDefs::iterator, i, attrs->attrs)
+        newEnv.vars[i->first] = i->second.displ = displ++;
+    
+    foreach (ExprAttrs::AttrDefs::iterator, i, attrs->attrs)
+        if (i->second.inherited) i->second.var.bind(env);
+        else i->second.e->bindVars(newEnv);
+    
+    body->bindVars(newEnv);
 }
 
-
-void checkVarDefs(const ATermMap & defs, Expr e)
+void ExprWith::bindVars(const StaticEnv & env)
 {
-    set<Expr> done;
-    checkVarDefs2(done, defs, e);
-}
-
-
-struct Canonicalise : TermFun
-{
-    ATerm operator () (ATerm e)
-    {
-        /* Remove position info. */
-        ATerm path;
-        int line, column;
-        if (matchPos(e, path, line, column))
-            return makeNoPos();
-
-        /* Sort attribute sets. */
-        ATermList _;
-        if (matchAttrs(e, _)) {
-            ATermMap attrs;
-            queryAllAttrs(e, attrs);
-            StringSet names;
-            for (ATermMap::const_iterator i = attrs.begin(); i != attrs.end(); ++i)
-                names.insert(aterm2String(i->key));
-
-            ATermList attrs2 = ATempty;
-            for (StringSet::reverse_iterator i = names.rbegin(); i != names.rend(); ++i)
-                attrs2 = ATinsert(attrs2,
-                    makeBind(toATerm(*i), attrs.get(toATerm(*i)), makeNoPos()));
-
-            return makeAttrs(attrs2);
+    /* Does this `with' have an enclosing `with'?  If so, record its
+       level so that `lookupVar' can look up variables in the previous
+       `with' if this one doesn't contain the desired attribute. */
+    const StaticEnv * curEnv;
+    unsigned int level;
+    prevWith = 0;
+    for (curEnv = &env, level = 1; curEnv; curEnv = curEnv->up, level++)
+        if (curEnv->isWith) {
+            prevWith = level;
+            break;
         }
-        
-        return e;
-    }
-};
+    
+    attrs->bindVars(env);    
+    StaticEnv newEnv(true, &env);
+    body->bindVars(newEnv);
+}
 
-
-Expr canonicaliseExpr(Expr e)
+void ExprIf::bindVars(const StaticEnv & env)
 {
-    Canonicalise canonicalise;
-    return bottomupRewrite(canonicalise, e);
+    cond->bindVars(env);
+    then->bindVars(env);
+    else_->bindVars(env);
 }
 
-
-Expr makeBool(bool b)
+void ExprAssert::bindVars(const StaticEnv & env)
 {
-    return b ? eTrue : eFalse;
+    cond->bindVars(env);
+    body->bindVars(env);
 }
 
-
-bool matchStr(Expr e, string & s, PathSet & context)
+void ExprOpNot::bindVars(const StaticEnv & env)
 {
-    ATermList l;
-    ATerm s_;
-
-    if (!matchStr(e, s_, l)) return false;
-
-    s = aterm2String(s_);
-
-    for (ATermIterator i(l); i; ++i)
-        context.insert(aterm2String(*i));
-
-    return true;
+    e->bindVars(env);
 }
 
-
-Expr makeStr(const string & s, const PathSet & context)
+void ExprConcatStrings::bindVars(const StaticEnv & env)
 {
-    return makeStr(toATerm(s), toATermList(context));
+    foreach (vector<Expr *>::iterator, i, *es)
+        (*i)->bindVars(env);
 }
 
 
-string showType(Expr e)
-{
-    ATerm t1, t2;
-    ATermList l1;
-    ATermBlob b1;
-    int i1;
-    Pattern p1;
-    if (matchStr(e, t1, l1)) return "a string";
-    if (matchPath(e, t1)) return "a path";
-    if (matchNull(e)) return "null";
-    if (matchInt(e, i1)) return "an integer";
-    if (matchBool(e, t1)) return "a boolean";
-    if (matchFunction(e, p1, t1, t2)) return "a function";
-    if (matchAttrs(e, l1)) return "an attribute set";
-    if (matchList(e, l1)) return "a list";
-    if (matchPrimOp(e, i1, b1, l1)) return "a partially applied built-in function";
-    return "an unknown type";
-}
-
-
-string showValue(Expr e)
-{
-    PathSet context;
-    string s;
-    ATerm s2;
-    int i;
-    if (matchStr(e, s, context)) {
-        string u;
-        for (string::iterator i = s.begin(); i != s.end(); ++i)
-            if (*i == '\"' || *i == '\\') u += "\\" + *i;
-            else if (*i == '\n') u += "\\n";
-            else if (*i == '\r') u += "\\r";
-            else if (*i == '\t') u += "\\t";
-            else u += *i;
-        return "\"" + u + "\"";
-    }
-    if (matchPath(e, s2)) return aterm2String(s2);
-    if (matchNull(e)) return "null";
-    if (matchInt(e, i)) return (format("%1%") % i).str();
-    if (e == eTrue) return "true";
-    if (e == eFalse) return "false";
-    /* !!! incomplete */
-    return "<unknown>";
-}
-
- 
 }

src/libexpr/nixexpr.hh

@@ -1,10 +1,9 @@
 #ifndef __NIXEXPR_H
 #define __NIXEXPR_H
 
-#include <map>
+#include "symbol-table.hh"
 
-#include "aterm-map.hh"
-#include "types.hh"
+#include <map>
 
 
 namespace nix {
@@ -16,107 +15,264 @@ MakeError(AssertionError, EvalError)
 MakeError(ThrownError, AssertionError)
 MakeError(Abort, EvalError)
 MakeError(TypeError, EvalError)
+MakeError(ImportError, EvalError) // error building an imported derivation
 
 
-/* Nix expressions are represented as ATerms.  The maximal sharing
-   property of the ATerm library allows us to implement caching of
-   normals forms efficiently. */
-typedef ATerm Expr;
-typedef ATerm DefaultValue;
-typedef ATerm Pos;
-typedef ATerm Pattern;
-typedef ATerm ATermBool;
+/* Position objects. */
 
-
-/* A STL vector of ATerms.  Should be used with great care since it's
-   stored on the heap, and the elements are therefore not roots to the
-   ATerm garbage collector. */
-typedef vector<ATerm> ATermVector;
-
-
-/* A substitution is a linked list of ATermMaps that map names to
-   identifiers.  We use a list of ATermMaps rather than a single to
-   make it easy to grow or shrink a substitution when entering a
-   scope. */
-struct Substitution
+struct Pos
 {
-    ATermMap * map;
-    const Substitution * prev;
+    string file;
+    unsigned int line, column;
+    Pos() : line(0), column(0) { };
+    Pos(const string & file, unsigned int line, unsigned int column)
+        : file(file), line(line), column(column) { };
+};
 
-    Substitution(const Substitution * prev, ATermMap * map)
-    {
-        this->prev = prev;
-        this->map = map;
-    }
+extern Pos noPos;
+
+std::ostream & operator << (std::ostream & str, const Pos & pos);
+
+
+struct Env;
+struct Value;
+struct EvalState;
+struct StaticEnv;
+
+
+/* Abstract syntax of Nix expressions. */
+
+struct Expr
+{
+    virtual void show(std::ostream & str);
+    virtual void bindVars(const StaticEnv & env);
+    virtual void eval(EvalState & state, Env & env, Value & v);
+};
+
+std::ostream & operator << (std::ostream & str, Expr & e);
+
+#define COMMON_METHODS \
+    void show(std::ostream & str); \
+    void eval(EvalState & state, Env & env, Value & v); \
+    void bindVars(const StaticEnv & env);
+
+struct ExprInt : Expr
+{
+    int n;
+    ExprInt(int n) : n(n) { };
+    COMMON_METHODS
+};
+
+struct ExprString : Expr
+{
+    Symbol s;
+    ExprString(const Symbol & s) : s(s) { };
+    COMMON_METHODS
+};
+
+/* Temporary class used during parsing of indented strings. */
+struct ExprIndStr : Expr
+{
+    string s;
+    ExprIndStr(const string & s) : s(s) { };
+};
+
+struct ExprPath : Expr
+{
+    string s;
+    ExprPath(const string & s) : s(s) { };
+    COMMON_METHODS
+};
+
+struct VarRef
+{
+    Symbol name;
+
+    /* Whether the variable comes from an environment (e.g. a rec, let
+       or function argument) or from a "with". */
+    bool fromWith;
     
-    Expr lookup(Expr name) const
-    {
-        Expr x;
-        for (const Substitution * s(this); s; s = s->prev)
-            if ((x = s->map->get(name))) return x;
-        return 0;
-    }
+    /* In the former case, the value is obtained by going `level'
+       levels up from the current environment and getting the
+       `displ'th value in that environment.  In the latter case, the
+       value is obtained by getting the attribute named `name' from
+       the attribute set stored in the environment that is `level'
+       levels up from the current one.*/
+    unsigned int level;
+    unsigned int displ;
+
+    VarRef() { };
+    VarRef(const Symbol & name) : name(name) { };
+    void bind(const StaticEnv & env);
 };
 
-
-/* Show a position. */
-string showPos(ATerm pos);
-
-/* Generic bottomup traversal over ATerms.  The traversal first
-   recursively descends into subterms, and then applies the given term
-   function to the resulting term. */
-struct TermFun
+struct ExprVar : Expr
 {
-    virtual ~TermFun() { }
-    virtual ATerm operator () (ATerm e) = 0;
+    VarRef info;
+    ExprVar(const Symbol & name) : info(name) { };
+    COMMON_METHODS
+};
+
+struct ExprSelect : Expr
+{
+    Expr * e;
+    Symbol name;
+    ExprSelect(Expr * e, const Symbol & name) : e(e), name(name) { };
+    COMMON_METHODS
+};
+
+struct ExprOpHasAttr : Expr
+{
+    Expr * e;
+    Symbol name;
+    ExprOpHasAttr(Expr * e, const Symbol & name) : e(e), name(name) { };
+    COMMON_METHODS
+};
+
+struct ExprAttrs : Expr
+{
+    bool recursive;
+    struct AttrDef {
+        bool inherited;
+        Expr * e; // if not inherited
+        VarRef var; // if inherited
+        Pos pos;
+        unsigned int displ; // displacement
+        AttrDef(Expr * e, const Pos & pos) : inherited(false), e(e), pos(pos) { };
+        AttrDef(const Symbol & name, const Pos & pos) : inherited(true), var(name), pos(pos) { };
+        AttrDef() { };
+    };
+    typedef std::map<Symbol, AttrDef> AttrDefs;
+    AttrDefs attrs;
+    ExprAttrs() : recursive(false) { };
+    COMMON_METHODS
+};
+
+struct ExprList : Expr
+{
+    std::vector<Expr *> elems;
+    ExprList() { };
+    COMMON_METHODS
+};
+
+struct Formal
+{
+    Symbol name;
+    Expr * def;
+    Formal(const Symbol & name, Expr * def) : name(name), def(def) { };
+};
+
+struct Formals
+{
+    typedef std::list<Formal> Formals_;
+    Formals_ formals;
+    std::set<Symbol> argNames; // used during parsing
+    bool ellipsis;
+};
+
+struct ExprLambda : Expr
+{
+    Pos pos;
+    Symbol arg;
+    bool matchAttrs;
+    Formals * formals;
+    Expr * body;
+    ExprLambda(const Pos & pos, const Symbol & arg, bool matchAttrs, Formals * formals, Expr * body)
+        : pos(pos), arg(arg), matchAttrs(matchAttrs), formals(formals), body(body)
+    {
+        if (!arg.empty() && formals && formals->argNames.find(arg) != formals->argNames.end())
+            throw ParseError(format("duplicate formal function argument `%1%' at %2%")
+                % arg % pos);
+    };
+    COMMON_METHODS
+};
+
+struct ExprLet : Expr
+{
+    ExprAttrs * attrs;
+    Expr * body;
+    ExprLet(ExprAttrs * attrs, Expr * body) : attrs(attrs), body(body) { };
+    COMMON_METHODS
+};
+
+struct ExprWith : Expr
+{
+    Pos pos;
+    Expr * attrs, * body;
+    unsigned int prevWith;
+    ExprWith(const Pos & pos, Expr * attrs, Expr * body) : pos(pos), attrs(attrs), body(body) { };
+    COMMON_METHODS
+};
+
+struct ExprIf : Expr
+{
+    Expr * cond, * then, * else_;
+    ExprIf(Expr * cond, Expr * then, Expr * else_) : cond(cond), then(then), else_(else_) { };
+    COMMON_METHODS
+};
+
+struct ExprAssert : Expr
+{
+    Pos pos;
+    Expr * cond, * body;
+    ExprAssert(const Pos & pos, Expr * cond, Expr * body) : pos(pos), cond(cond), body(body) { };
+    COMMON_METHODS
+};
+
+struct ExprOpNot : Expr
+{
+    Expr * e;
+    ExprOpNot(Expr * e) : e(e) { };
+    COMMON_METHODS
+};
+
+#define MakeBinOp(name, s) \
+    struct Expr##name : Expr \
+    { \
+        Expr * e1, * e2; \
+        Expr##name(Expr * e1, Expr * e2) : e1(e1), e2(e2) { }; \
+        void show(std::ostream & str) \
+        { \
+            str << *e1 << " " s " " << *e2; \
+        } \
+        void bindVars(const StaticEnv & env) \
+        { \
+            e1->bindVars(env); e2->bindVars(env); \
+        } \
+        void eval(EvalState & state, Env & env, Value & v); \
+    };
+
+MakeBinOp(App, "")
+MakeBinOp(OpEq, "==")
+MakeBinOp(OpNEq, "!=")
+MakeBinOp(OpAnd, "&&")
+MakeBinOp(OpOr, "||")
+MakeBinOp(OpImpl, "->")
+MakeBinOp(OpUpdate, "//")
+MakeBinOp(OpConcatLists, "++")
+
+struct ExprConcatStrings : Expr
+{
+    vector<Expr *> * es;
+    ExprConcatStrings(vector<Expr *> * es) : es(es) { };
+    COMMON_METHODS
 };
-ATerm bottomupRewrite(TermFun & f, ATerm e);
 
 
-/* Query all attributes in an attribute set expression.  The
-   expression must be in normal form. */
-void queryAllAttrs(Expr e, ATermMap & attrs, bool withPos = false);
-
-/* Query a specific attribute from an attribute set expression.  The
-   expression must be in normal form. */
-Expr queryAttr(Expr e, const string & name);
-Expr queryAttr(Expr e, const string & name, ATerm & pos);
-
-/* Create an attribute set expression from an Attrs value. */
-Expr makeAttrs(const ATermMap & attrs);
+/* Static environments are used to map variable names onto (level,
+   displacement) pairs used to obtain the value of the variable at
+   runtime. */
+struct StaticEnv
+{
+    bool isWith;
+    const StaticEnv * up;
+    typedef std::map<Symbol, unsigned int> Vars;
+    Vars vars;
+    StaticEnv(bool isWith, const StaticEnv * up) : isWith(isWith), up(up) { };
+};
 
 
-/* Perform a set of substitutions on an expression. */
-Expr substitute(const Substitution & subs, Expr e);
 
-
-/* Check whether all variables are defined in the given expression.
-   Throw an exception if this isn't the case. */
-void checkVarDefs(const ATermMap & def, Expr e);
-
-
-/* Canonicalise a Nix expression by sorting attributes and removing
-   location information. */
-Expr canonicaliseExpr(Expr e);
-
-
-/* Create an expression representing a boolean. */
-Expr makeBool(bool b);
-
-
-/* Manipulation of Str() nodes.  Note: matchStr() does not clear
-   context!  */
-bool matchStr(Expr e, string & s, PathSet & context);
-
-Expr makeStr(const string & s, const PathSet & context = PathSet());
-
-
-/* Showing types, values. */
-string showType(Expr e);
-
-string showValue(Expr e);
-
- 
 }
 
 

src/libexpr/parser.hh

@@ -8,12 +8,11 @@ namespace nix {
 
 
 /* Parse a Nix expression from the specified file.  If `path' refers
-   to a directory, the "/default.nix" is appended. */
-Expr parseExprFromFile(EvalState & state, Path path);
+   to a directory, then "/default.nix" is appended. */
+Expr * parseExprFromFile(EvalState & state, Path path);
 
 /* Parse a Nix expression from the specified string. */
-Expr parseExprFromString(EvalState & state, const string & s,
-    const Path & basePath);
+Expr * parseExprFromString(EvalState & state, const string & s, const Path & basePath);
 
 
 }

src/libexpr/parser.y

@@ -7,171 +7,126 @@
 %parse-param { yyscan_t scanner }
 %parse-param { ParseData * data }
 %lex-param { yyscan_t scanner }
+%lex-param { ParseData * data }
 
-
-%{
-/* Newer versions of Bison copy the declarations below to
-   parser-tab.hh, which sucks bigtime since lexer.l doesn't want that
-   stuff.  So allow it to be excluded. */
-#ifndef BISON_HEADER_HACK
-#define BISON_HEADER_HACK
+%code requires {
+    
+#ifndef BISON_HEADER
+#define BISON_HEADER
     
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-
-#include "aterm.hh"
 #include "util.hh"
     
+#include "nixexpr.hh"
+
+namespace nix {
+
+    struct ParseData 
+    {
+        SymbolTable & symbols;
+        Expr * result;
+        Path basePath;
+        Path path;
+        string error;
+        Symbol sLetBody;
+        ParseData(SymbolTable & symbols)
+            : symbols(symbols)
+            , sLetBody(symbols.create("<let-body>"))
+            { };
+    };
+    
+}
+
+#define YY_DECL int yylex \
+    (YYSTYPE * yylval_param, YYLTYPE * yylloc_param, yyscan_t yyscanner, nix::ParseData * data)
+
+#endif
+
+}
+
+%{
+
 #include "parser-tab.hh"
 #include "lexer-tab.hh"
 #define YYSTYPE YYSTYPE // workaround a bug in Bison 2.4
 
-#include "nixexpr.hh"
-#include "nixexpr-ast.hh"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
 
+YY_DECL;
 
 using namespace nix;
 
 
 namespace nix {
-
     
-struct ParseData 
-{
-    Expr result;
-    Path basePath;
-    Path path;
-    string error;
-};
- 
 
-static string showAttrPath(ATermList attrPath)
+static string showAttrPath(const vector<Symbol> & attrPath)
 {
     string s;
-    for (ATermIterator i(attrPath); i; ++i) {
+    foreach (vector<Symbol>::const_iterator, i, attrPath) {
         if (!s.empty()) s += '.';
-        s += aterm2String(*i);
+        s += *i;
     }
     return s;
 }
- 
 
-struct Tree
+
+static void dupAttr(const vector<Symbol> & attrPath, const Pos & pos, const Pos & prevPos)
 {
-    Expr leaf; ATerm pos; bool recursive;
-    typedef std::map<ATerm, Tree> Children;
-    Children children;
-    Tree() { leaf = 0; recursive = true; }
-};
-
-
-static ATermList buildAttrs(const Tree & t, ATermList & nonrec)
-{
-    ATermList res = ATempty;
-    for (Tree::Children::const_reverse_iterator i = t.children.rbegin();
-         i != t.children.rend(); ++i)
-        if (!i->second.recursive)
-            nonrec = ATinsert(nonrec, makeBind(i->first, i->second.leaf, i->second.pos));
-        else
-            res = ATinsert(res, i->second.leaf
-                ? makeBind(i->first, i->second.leaf, i->second.pos)
-                : makeBind(i->first, makeAttrs(buildAttrs(i->second, nonrec)), makeNoPos()));
-    return res;
+    throw ParseError(format("attribute `%1%' at %2% already defined at %3%")
+        % showAttrPath(attrPath) % pos % prevPos);
 }
  
 
-static Expr fixAttrs(bool recursive, ATermList as)
+static void dupAttr(Symbol attr, const Pos & pos, const Pos & prevPos)
 {
-    Tree attrs;
+    vector<Symbol> attrPath; attrPath.push_back(attr);
+    throw ParseError(format("attribute `%1%' at %2% already defined at %3%")
+        % showAttrPath(attrPath) % pos % prevPos);
+}
+ 
 
-    for (ATermIterator i(as); i; ++i) {
-        ATermList names, attrPath; Expr src, e; ATerm name, pos;
-
-        if (matchInherit(*i, src, names, pos)) {
-            bool fromScope = matchScope(src);
-            for (ATermIterator j(names); j; ++j) {
-                if (attrs.children.find(*j) != attrs.children.end()) 
-                    throw ParseError(format("duplicate definition of attribute `%1%' at %2%")
-                        % showAttrPath(ATmakeList1(*j)) % showPos(pos));
-                Tree & t(attrs.children[*j]);
-                t.leaf = fromScope ? makeVar(*j) : makeSelect(src, *j);
-                t.pos = pos;
-                if (recursive && fromScope) t.recursive = false;
+static void addAttr(ExprAttrs * attrs, const vector<Symbol> & attrPath,
+    Expr * e, const Pos & pos)
+{
+    unsigned int n = 0;
+    foreach (vector<Symbol>::const_iterator, i, attrPath) {
+        n++;
+        ExprAttrs::AttrDefs::iterator j = attrs->attrs.find(*i);
+        if (j != attrs->attrs.end()) {
+            if (!j->second.inherited) {
+                ExprAttrs * attrs2 = dynamic_cast<ExprAttrs *>(j->second.e);
+                if (!attrs2 || n == attrPath.size()) dupAttr(attrPath, pos, j->second.pos);
+                attrs = attrs2;
+            } else
+                dupAttr(attrPath, pos, j->second.pos);
+        } else {
+            if (n == attrPath.size())
+                attrs->attrs[*i] = ExprAttrs::AttrDef(e, pos);
+            else {
+                ExprAttrs * nested = new ExprAttrs;
+                attrs->attrs[*i] = ExprAttrs::AttrDef(nested, pos);
+                attrs = nested;
             }
         }
-
-        else if (matchBindAttrPath(*i, attrPath, e, pos)) {
-
-            Tree * t(&attrs);
-            
-            for (ATermIterator j(attrPath); j; ) {
-                name = *j; ++j;
-                if (t->leaf) throw ParseError(format("attribute set containing `%1%' at %2% already defined at %3%")
-                    % showAttrPath(attrPath) % showPos(pos) % showPos (t->pos));
-                t = &(t->children[name]);
-            }
-
-            if (t->leaf)
-                throw ParseError(format("duplicate definition of attribute `%1%' at %2% and %3%")
-                    % showAttrPath(attrPath) % showPos(pos) % showPos (t->pos));
-            if (!t->children.empty())
-                throw ParseError(format("duplicate definition of attribute `%1%' at %2%")
-                    % showAttrPath(attrPath) % showPos(pos));
-
-            t->leaf = e; t->pos = pos;
-        }
-
-        else abort(); /* can't happen */
     }
-
-    ATermList nonrec = ATempty;
-    ATermList rec = buildAttrs(attrs, nonrec);
-        
-    return recursive ? makeRec(rec, nonrec) : makeAttrs(rec);
 }
 
 
-static void checkPatternVars(ATerm pos, ATermMap & map, Pattern pat)
+static void addFormal(const Pos & pos, Formals * formals, const Formal & formal)
 {
-    ATerm name;
-    ATermList formals;
-    Pattern pat1, pat2;
-    ATermBool ellipsis;
-    if (matchVarPat(pat, name)) {
-        if (map.get(name))
-            throw ParseError(format("duplicate formal function argument `%1%' at %2%")
-                % aterm2String(name) % showPos(pos));
-        map.set(name, name);
-    }
-    else if (matchAttrsPat(pat, formals, ellipsis)) { 
-        for (ATermIterator i(formals); i; ++i) {
-            ATerm d1;
-            if (!matchFormal(*i, name, d1)) abort();
-            if (map.get(name))
-                throw ParseError(format("duplicate formal function argument `%1%' at %2%")
-                    % aterm2String(name) % showPos(pos));
-            map.set(name, name);
-        }
-    }
-    else if (matchAtPat(pat, pat1, pat2)) {
-        checkPatternVars(pos, map, pat1);
-        checkPatternVars(pos, map, pat2);
-    }
-    else abort();
+    if (formals->argNames.find(formal.name) != formals->argNames.end())
+        throw ParseError(format("duplicate formal function argument `%1%' at %2%")
+            % formal.name % pos);
+    formals->formals.push_front(formal);
+    formals->argNames.insert(formal.name);
 }
 
 
-static void checkPatternVars(ATerm pos, Pattern pat)
+static Expr * stripIndentation(SymbolTable & symbols, vector<Expr *> & es)
 {
-    ATermMap map;
-    checkPatternVars(pos, map, pat);
-}
-
-
-static Expr stripIndentation(ATermList es)
-{
-    if (es == ATempty) return makeStr("");
+    if (es.empty()) return new ExprString(symbols.create(""));
     
     /* Figure out the minimum indentation.  Note that by design
        whitespace-only final lines are not taken into account.  (So
@@ -179,9 +134,9 @@ static Expr stripIndentation(ATermList es)
     bool atStartOfLine = true; /* = seen only whitespace in the current line */
     unsigned int minIndent = 1000000;
     unsigned int curIndent = 0;
-    ATerm e;
-    for (ATermIterator i(es); i; ++i) {
-        if (!matchIndStr(*i, e)) {
+    foreach (vector<Expr *>::iterator, i, es) {
+        ExprIndStr * e = dynamic_cast<ExprIndStr *>(*i);
+        if (!e) {
             /* Anti-quotations end the current start-of-line whitespace. */
             if (atStartOfLine) {
                 atStartOfLine = false;
@@ -189,12 +144,11 @@ static Expr stripIndentation(ATermList es)
             }
             continue;
         }
-        string s = aterm2String(e);
-        for (unsigned int j = 0; j < s.size(); ++j) {
+        for (unsigned int j = 0; j < e->s.size(); ++j) {
             if (atStartOfLine) {
-                if (s[j] == ' ')
+                if (e->s[j] == ' ')
                     curIndent++;
-                else if (s[j] == '\n') {
+                else if (e->s[j] == '\n') {
                     /* Empty line, doesn't influence minimum
                        indentation. */
                     curIndent = 0;
@@ -202,7 +156,7 @@ static Expr stripIndentation(ATermList es)
                     atStartOfLine = false;
                     if (curIndent < minIndent) minIndent = curIndent;
                 }
-            } else if (s[j] == '\n') {
+            } else if (e->s[j] == '\n') {
                 atStartOfLine = true;
                 curIndent = 0;
             }
@@ -210,37 +164,37 @@ static Expr stripIndentation(ATermList es)
     }
 
     /* Strip spaces from each line. */
-    ATermList es2 = ATempty;
+    vector<Expr *> * es2 = new vector<Expr *>;
     atStartOfLine = true;
     unsigned int curDropped = 0;
-    unsigned int n = ATgetLength(es);
-    for (ATermIterator i(es); i; ++i, --n) {
-        if (!matchIndStr(*i, e)) {
+    unsigned int n = es.size();
+    for (vector<Expr *>::iterator i = es.begin(); i != es.end(); ++i, --n) {
+        ExprIndStr * e = dynamic_cast<ExprIndStr *>(*i);
+        if (!e) {
             atStartOfLine = false;
             curDropped = 0;
-            es2 = ATinsert(es2, *i);
+            es2->push_back(*i);
             continue;
         }
         
-        string s = aterm2String(e);
         string s2;
-        for (unsigned int j = 0; j < s.size(); ++j) {
+        for (unsigned int j = 0; j < e->s.size(); ++j) {
             if (atStartOfLine) {
-                if (s[j] == ' ') {
+                if (e->s[j] == ' ') {
                     if (curDropped++ >= minIndent)
-                        s2 += s[j];
+                        s2 += e->s[j];
                 }
-                else if (s[j] == '\n') {
+                else if (e->s[j] == '\n') {
                     curDropped = 0;
-                    s2 += s[j];
+                    s2 += e->s[j];
                 } else {
                     atStartOfLine = false;
                     curDropped = 0;
-                    s2 += s[j];
+                    s2 += e->s[j];
                 }
             } else {
-                s2 += s[j];
-                if (s[j] == '\n') atStartOfLine = true;
+                s2 += e->s[j];
+                if (e->s[j] == '\n') atStartOfLine = true;
             }
         }
 
@@ -251,11 +205,11 @@ static Expr stripIndentation(ATermList es)
             if (p != string::npos && s2.find_first_not_of(' ', p + 1) == string::npos)
                 s2 = string(s2, 0, p + 1);
         }
-            
-        es2 = ATinsert(es2, makeStr(s2));
+
+        es2->push_back(new ExprString(symbols.create(s2)));
     }
 
-    return makeConcatStrings(ATreverse(es2));
+    return new ExprConcatStrings(es2);
 }
 
 
@@ -263,13 +217,12 @@ void backToString(yyscan_t scanner);
 void backToIndString(yyscan_t scanner);
 
 
-static Pos makeCurPos(YYLTYPE * loc, ParseData * data)
+static Pos makeCurPos(const YYLTYPE & loc, ParseData * data)
 {
-    return makePos(toATerm(data->path),
-        loc->first_line, loc->first_column);
+    return Pos(data->path, loc.first_line, loc.first_column);
 }
 
-#define CUR_POS makeCurPos(yylocp, data)
+#define CUR_POS makeCurPos(*yylocp, data)
 
 
 }
@@ -277,50 +230,40 @@ static Pos makeCurPos(YYLTYPE * loc, ParseData * data)
 
 void yyerror(YYLTYPE * loc, yyscan_t scanner, ParseData * data, const char * error)
 {
-    data->error = (format("%1%, at `%2%':%3%:%4%")
-        % error % data->path % loc->first_line % loc->first_column).str();
+    data->error = (format("%1%, at %2%")
+        % error % makeCurPos(*loc, data)).str();
 }
 
 
-/* Make sure that the parse stack is scanned by the ATerm garbage
-   collector. */
-static void * mallocAndProtect(size_t size)
-{
-    void * p = malloc(size);
-    if (p) ATprotectMemory(p, size);
-    return p;
-}
-
-static void freeAndUnprotect(void * p)
-{
-    ATunprotectMemory(p);
-    free(p);
-}
-
-#define YYMALLOC mallocAndProtect
-#define YYFREE freeAndUnprotect
-
-
-#endif
-
-
 %}
 
 %union {
-  ATerm t;
-  ATermList ts;
-  struct {
-    ATermList formals;
-    bool ellipsis;
-  } formals;
+  nix::Expr * e;
+  nix::ExprList * list;
+  nix::ExprAttrs * attrs;
+  nix::Formals * formals;
+  nix::Formal * formal;
+  int n;
+  char * id; // !!! -> Symbol
+  char * path;
+  char * uri;
+  std::vector<nix::Symbol> * ids;
+  std::vector<nix::Expr *> * string_parts;
 }
 
-%type <t> start expr expr_function expr_if expr_op
-%type <t> expr_app expr_select expr_simple bind inheritsrc formal
-%type <t> pattern pattern2
-%type <ts> binds ids attrpath expr_list string_parts ind_string_parts
+%type <e> start expr expr_function expr_if expr_op
+%type <e> expr_app expr_select expr_simple
+%type <list> expr_list
+%type <attrs> binds
 %type <formals> formals
-%token <t> ID INT STR IND_STR PATH URI
+%type <formal> formal
+%type <ids> ids attrpath
+%type <string_parts> string_parts ind_string_parts
+%token <id> ID ATTRPATH
+%token <e> STR IND_STR
+%token <n> INT
+%token <path> PATH
+%token <uri> URI
 %token IF THEN ELSE ASSERT WITH LET IN REC INHERIT EQ NEQ AND OR IMPL
 %token DOLLAR_CURLY /* == ${ */
 %token IND_STRING_OPEN IND_STRING_CLOSE
@@ -344,163 +287,170 @@ start: expr { data->result = $1; };
 expr: expr_function;
 
 expr_function
-  : pattern ':' expr_function
-    { checkPatternVars(CUR_POS, $1); $$ = makeFunction($1, $3, CUR_POS); }
+  : ID ':' expr_function
+    { $$ = new ExprLambda(CUR_POS, data->symbols.create($1), false, 0, $3); }
+  | '{' formals '}' ':' expr_function
+    { $$ = new ExprLambda(CUR_POS, data->symbols.create(""), true, $2, $5); }
+  | '{' formals '}' '@' ID ':' expr_function
+    { $$ = new ExprLambda(CUR_POS, data->symbols.create($5), true, $2, $7); }
+  | ID '@' '{' formals '}' ':' expr_function
+    { $$ = new ExprLambda(CUR_POS, data->symbols.create($1), true, $4, $7); }
   | ASSERT expr ';' expr_function
-    { $$ = makeAssert($2, $4, CUR_POS); }
+    { $$ = new ExprAssert(CUR_POS, $2, $4); }
   | WITH expr ';' expr_function
-    { $$ = makeWith($2, $4, CUR_POS); }
+    { $$ = new ExprWith(CUR_POS, $2, $4); }
   | LET binds IN expr_function
-    { $$ = makeSelect(fixAttrs(true, ATinsert($2, makeBindAttrPath(ATmakeList1(toATerm("<let-body>")), $4, CUR_POS))), toATerm("<let-body>")); }
+    { $$ = new ExprLet($2, $4); }
   | expr_if
   ;
 
 expr_if
-  : IF expr THEN expr ELSE expr
-    { $$ = makeIf($2, $4, $6); }
+  : IF expr THEN expr ELSE expr { $$ = new ExprIf($2, $4, $6); }
   | expr_op
   ;
 
 expr_op
-  : '!' expr_op %prec NEG { $$ = makeOpNot($2); }
-  | expr_op EQ expr_op { $$ = makeOpEq($1, $3); }
-  | expr_op NEQ expr_op { $$ = makeOpNEq($1, $3); }
-  | expr_op AND expr_op { $$ = makeOpAnd($1, $3); }
-  | expr_op OR expr_op { $$ = makeOpOr($1, $3); }
-  | expr_op IMPL expr_op { $$ = makeOpImpl($1, $3); }
-  | expr_op UPDATE expr_op { $$ = makeOpUpdate($1, $3); }
-  | expr_op '~' expr_op { $$ = makeSubPath($1, $3); }
-  | expr_op '?' ID { $$ = makeOpHasAttr($1, $3); }
-  | expr_op '+' expr_op { $$ = makeOpPlus($1, $3); }
-  | expr_op CONCAT expr_op { $$ = makeOpConcat($1, $3); }
+  : '!' expr_op %prec NEG { $$ = new ExprOpNot($2); }
+  | expr_op EQ expr_op { $$ = new ExprOpEq($1, $3); }
+  | expr_op NEQ expr_op { $$ = new ExprOpNEq($1, $3); }
+  | expr_op AND expr_op { $$ = new ExprOpAnd($1, $3); }
+  | expr_op OR expr_op { $$ = new ExprOpOr($1, $3); }
+  | expr_op IMPL expr_op { $$ = new ExprOpImpl($1, $3); }
+  | expr_op UPDATE expr_op { $$ = new ExprOpUpdate($1, $3); }
+  | expr_op '?' ID { $$ = new ExprOpHasAttr($1, data->symbols.create($3)); }
+  | expr_op '+' expr_op
+    { vector<Expr *> * l = new vector<Expr *>;
+      l->push_back($1);
+      l->push_back($3);
+      $$ = new ExprConcatStrings(l);
+    }
+  | expr_op CONCAT expr_op { $$ = new ExprOpConcatLists($1, $3); }
   | expr_app
   ;
 
 expr_app
   : expr_app expr_select
-    { $$ = makeCall($1, $2); }
+    { $$ = new ExprApp($1, $2); }
   | expr_select { $$ = $1; }
   ;
 
 expr_select
   : expr_select '.' ID
-    { $$ = makeSelect($1, $3); }
+    { $$ = new ExprSelect($1, data->symbols.create($3)); }
   | expr_simple { $$ = $1; }
   ;
 
 expr_simple
-  : ID { $$ = makeVar($1); }
-  | INT { $$ = makeInt(ATgetInt((ATermInt) $1)); }
+  : ID { $$ = new ExprVar(data->symbols.create($1)); }
+  | INT { $$ = new ExprInt($1); }
   | '"' string_parts '"' {
       /* For efficiency, and to simplify parse trees a bit. */
-      if ($2 == ATempty) $$ = makeStr(toATerm(""), ATempty);
-      else if (ATgetNext($2) == ATempty) $$ = ATgetFirst($2);
-      else $$ = makeConcatStrings(ATreverse($2));
+      if ($2->empty()) $$ = new ExprString(data->symbols.create(""));
+      else if ($2->size() == 1) $$ = $2->front();
+      else $$ = new ExprConcatStrings($2);
   }
   | IND_STRING_OPEN ind_string_parts IND_STRING_CLOSE {
-      $$ = stripIndentation(ATreverse($2));
+      $$ = stripIndentation(data->symbols, *$2);
   }
-  | PATH { $$ = makePath(toATerm(absPath(aterm2String($1), data->basePath))); }
-  | URI { $$ = makeStr($1, ATempty); }
+  | PATH { $$ = new ExprPath(absPath($1, data->basePath)); }
+  | URI { $$ = new ExprString(data->symbols.create($1)); }
   | '(' expr ')' { $$ = $2; }
   /* Let expressions `let {..., body = ...}' are just desugared
      into `(rec {..., body = ...}).body'. */
   | LET '{' binds '}'
-    { $$ = makeSelect(fixAttrs(true, $3), toATerm("body")); }
+    { $3->recursive = true; $$ = new ExprSelect($3, data->symbols.create("body")); }
   | REC '{' binds '}'
-    { $$ = fixAttrs(true, $3); }
+    { $3->recursive = true; $$ = $3; }
   | '{' binds '}'
-    { $$ = fixAttrs(false, $2); }
-  | '[' expr_list ']' { $$ = makeList(ATreverse($2)); }
+    { $$ = $2; }
+  | '[' expr_list ']' { $$ = $2; }
   ;
 
 string_parts
-  : string_parts STR { $$ = ATinsert($1, $2); }
-  | string_parts DOLLAR_CURLY expr '}' { backToString(scanner); $$ = ATinsert($1, $3); }
-  | { $$ = ATempty; }
+  : string_parts STR { $$ = $1; $1->push_back($2); }
+  | string_parts DOLLAR_CURLY expr '}' { backToString(scanner); $$ = $1; $1->push_back($3); }
+  | { $$ = new vector<Expr *>; }
   ;
 
 ind_string_parts
-  : ind_string_parts IND_STR { $$ = ATinsert($1, $2); }
-  | ind_string_parts DOLLAR_CURLY expr '}' { backToIndString(scanner); $$ = ATinsert($1, $3); }
-  | { $$ = ATempty; }
-  ;
-
-pattern
-  : pattern2 '@' pattern { $$ = makeAtPat($1, $3); }
-  | pattern2
-  ;
-
-pattern2
-  : ID { $$ = makeVarPat($1); }
-  | '{' formals '}' { $$ = makeAttrsPat($2.formals, $2.ellipsis ? eTrue : eFalse); }
+  : ind_string_parts IND_STR { $$ = $1; $1->push_back($2); }
+  | ind_string_parts DOLLAR_CURLY expr '}' { backToIndString(scanner); $$ = $1; $1->push_back($3); }
+  | { $$ = new vector<Expr *>; }
   ;
 
 binds
-  : binds bind { $$ = ATinsert($1, $2); }
-  | { $$ = ATempty; }
+  : binds attrpath '=' expr ';' { $$ = $1; addAttr($$, *$2, $4, makeCurPos(@2, data)); }
+  | binds INHERIT ids ';'
+    { $$ = $1;
+      foreach (vector<Symbol>::iterator, i, *$3) {
+          if ($$->attrs.find(*i) != $$->attrs.end())
+              dupAttr(*i, makeCurPos(@3, data), $$->attrs[*i].pos);
+          Pos pos = makeCurPos(@3, data);
+          $$->attrs[*i] = ExprAttrs::AttrDef(*i, pos);
+      }
+    }
+  | binds INHERIT '(' expr ')' ids ';'
+    { $$ = $1;
+      /* !!! Should ensure sharing of the expression in $4. */
+      foreach (vector<Symbol>::iterator, i, *$6) {
+          if ($$->attrs.find(*i) != $$->attrs.end())
+              dupAttr(*i, makeCurPos(@6, data), $$->attrs[*i].pos);
+          $$->attrs[*i] = ExprAttrs::AttrDef(new ExprSelect($4, *i), makeCurPos(@6, data));
+      }}
+
+  | { $$ = new ExprAttrs; }
   ;
 
-bind
-  : attrpath '=' expr ';'
-    { $$ = makeBindAttrPath(ATreverse($1), $3, CUR_POS); }
-  | INHERIT inheritsrc ids ';'
-    { $$ = makeInherit($2, $3, CUR_POS); }
+ids
+  : ids ID { $$ = $1; $1->push_back(data->symbols.create($2)); /* !!! dangerous */ }
+  | { $$ = new vector<Symbol>; }
   ;
 
-inheritsrc
-  : '(' expr ')' { $$ = $2; }
-  | { $$ = makeScope(); }
-  ;
-
-ids: ids ID { $$ = ATinsert($1, $2); } | { $$ = ATempty; };
-
 attrpath
-  : attrpath '.' ID { $$ = ATinsert($1, $3); }
-  | ID { $$ = ATmakeList1($1); }
+  : attrpath '.' ID { $$ = $1; $1->push_back(data->symbols.create($3)); }
+  | ID { $$ = new vector<Symbol>; $$->push_back(data->symbols.create($1)); }
   ;
 
 expr_list
-  : expr_list expr_select { $$ = ATinsert($1, $2); }
-  | { $$ = ATempty; }
+  : expr_list expr_select { $$ = $1; $1->elems.push_back($2); /* !!! dangerous */ }
+  | { $$ = new ExprList; }
   ;
 
 formals
-  : formal ',' formals /* !!! right recursive */
-    { $$.formals = ATinsert($3.formals, $1); $$.ellipsis = $3.ellipsis; }
+  : formal ',' formals
+    { $$ = $3; addFormal(CUR_POS, $$, *$1); }
   | formal
-    { $$.formals = ATinsert(ATempty, $1); $$.ellipsis = false; }
+    { $$ = new Formals; addFormal(CUR_POS, $$, *$1); $$->ellipsis = false; }
   |
-    { $$.formals = ATempty; $$.ellipsis = false; }
+    { $$ = new Formals; $$->ellipsis = false; }
   | ELLIPSIS
-    { $$.formals = ATempty; $$.ellipsis = true; }
+    { $$ = new Formals; $$->ellipsis = true; }
   ;
 
 formal
-  : ID { $$ = makeFormal($1, makeNoDefaultValue()); }
-  | ID '?' expr { $$ = makeFormal($1, makeDefaultValue($3)); }
+  : ID { $$ = new Formal(data->symbols.create($1), 0); }
+  | ID '?' expr { $$ = new Formal(data->symbols.create($1), $3); }
   ;
   
 %%
 
 
-#include "eval.hh"  
-
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <fcntl.h>
 #include <unistd.h>
 
+#include <eval.hh>
+
 
 namespace nix {
       
 
-static Expr parse(EvalState & state,
-    const char * text, const Path & path,
-    const Path & basePath)
+static Expr * parse(EvalState & state, const char * text,
+    const Path & path, const Path & basePath)
 {
     yyscan_t scanner;
-    ParseData data;
+    ParseData data(state.symbols);
     data.basePath = basePath;
     data.path = path;
 
@@ -512,7 +462,7 @@ static Expr parse(EvalState & state,
     if (res) throw ParseError(data.error);
 
     try {
-        checkVarDefs(state.primOps, data.result);
+        data.result->bindVars(state.staticBaseEnv);
     } catch (Error & e) {
         throw ParseError(format("%1%, in `%2%'") % e.msg() % path);
     }
@@ -521,16 +471,10 @@ static Expr parse(EvalState & state,
 }
 
 
-Expr parseExprFromFile(EvalState & state, Path path)
+Expr * parseExprFromFile(EvalState & state, Path path)
 {
     assert(path[0] == '/');
 
-#if 0
-    /* Perhaps this is already an imploded parse tree? */
-    Expr e = ATreadFromNamedFile(path.c_str());
-    if (e) return e;
-#endif
-
     /* If `path' is a symlink, follow it.  This is so that relative
        path references work. */
     struct stat st;
@@ -542,8 +486,6 @@ Expr parseExprFromFile(EvalState & state, Path path)
     }
 
     /* If `path' refers to a directory, append `/default.nix'. */
-    if (stat(path.c_str(), &st))
-        throw SysError(format("getting status of `%1%'") % path);
     if (S_ISDIR(st.st_mode))
         path = canonPath(path + "/default.nix");
 
@@ -552,7 +494,7 @@ Expr parseExprFromFile(EvalState & state, Path path)
 }
 
 
-Expr parseExprFromString(EvalState & state,
+Expr * parseExprFromString(EvalState & state,
     const string & s, const Path & basePath)
 {
     return parse(state, s.c_str(), "(string)", basePath);

src/libexpr/symbol-table.hh

@@ -0,0 +1,92 @@
+#ifndef __SYMBOL_TABLE_H
+#define __SYMBOL_TABLE_H
+
+#include "config.h"
+
+#include <map>
+
+#if HAVE_TR1_UNORDERED_SET
+#include <tr1/unordered_set>
+#endif
+
+#include "types.hh"
+
+namespace nix {
+
+/* Symbol table used by the parser and evaluator to represent and look
+   up identifiers and attribute sets efficiently.
+   SymbolTable::create() converts a string into a symbol.  Symbols
+   have the property that they can be compared efficiently (using a
+   pointer equality test), because the symbol table stores only one
+   copy of each string. */
+
+class Symbol
+{
+private:
+    const string * s; // pointer into SymbolTable
+    Symbol(const string * s) : s(s) { };
+    friend class SymbolTable;
+
+public:
+    Symbol() : s(0) { };
+    
+    bool operator == (const Symbol & s2) const
+    {
+        return s == s2.s;
+    }
+    
+    bool operator != (const Symbol & s2) const
+    {
+        return s != s2.s;
+    }
+    
+    bool operator < (const Symbol & s2) const
+    {
+        return s < s2.s;
+    }
+
+    operator const string & () const
+    {
+        return *s;
+    }
+
+    bool empty() const
+    {
+        return s->empty();
+    }
+
+    friend std::ostream & operator << (std::ostream & str, const Symbol & sym);
+};
+
+inline std::ostream & operator << (std::ostream & str, const Symbol & sym)
+{
+    str << *sym.s;
+    return str;
+}
+
+class SymbolTable
+{
+private:
+#if HAVE_TR1_UNORDERED_SET 
+    typedef std::tr1::unordered_set<string> Symbols;
+#else
+    typedef std::set<string> Symbols;
+#endif
+    Symbols symbols;
+
+public:
+    Symbol create(const string & s)
+    {
+        std::pair<Symbols::iterator, bool> res = symbols.insert(s);
+        return Symbol(&*res.first);
+    }
+
+    unsigned int size() const
+    {
+        return symbols.size();
+    }
+};
+
+}
+
+#endif /* !__SYMBOL_TABLE_H */

src/libexpr/value-to-xml.cc

@@ -0,0 +1,162 @@
+#include "value-to-xml.hh"
+#include "xml-writer.hh"
+#include "util.hh"
+
+#include <cstdlib>
+
+
+namespace nix {
+
+    
+static XMLAttrs singletonAttrs(const string & name, const string & value)
+{
+    XMLAttrs attrs;
+    attrs[name] = value;
+    return attrs;
+}
+
+
+static void printValueAsXML(EvalState & state, bool strict, bool location,
+    Value & v, XMLWriter & doc, PathSet & context, PathSet & drvsSeen);
+
+
+static void posToXML(XMLAttrs & xmlAttrs, const Pos & pos)
+{
+    xmlAttrs["path"] = pos.file;
+    xmlAttrs["line"] = (format("%1%") % pos.line).str();
+    xmlAttrs["column"] = (format("%1%") % pos.column).str();
+}
+
+
+static void showAttrs(EvalState & state, bool strict, bool location,
+    Bindings & attrs, XMLWriter & doc, PathSet & context, PathSet & drvsSeen)
+{
+    StringSet names;
+    
+    foreach (Bindings::iterator, i, attrs)
+        names.insert(i->name);
+    
+    foreach (StringSet::iterator, i, names) {
+        Attr & a(*attrs.find(state.symbols.create(*i)));
+        
+        XMLAttrs xmlAttrs;
+        xmlAttrs["name"] = *i;
+        if (location && a.pos != &noPos) posToXML(xmlAttrs, *a.pos);
+        
+        XMLOpenElement _(doc, "attr", xmlAttrs);
+        printValueAsXML(state, strict, location,
+            *a.value, doc, context, drvsSeen);
+    }
+}
+
+
+static void printValueAsXML(EvalState & state, bool strict, bool location,
+    Value & v, XMLWriter & doc, PathSet & context, PathSet & drvsSeen)
+{
+    checkInterrupt();
+
+    if (strict) state.forceValue(v);
+        
+    switch (v.type) {
+
+        case tInt:
+            doc.writeEmptyElement("int", singletonAttrs("value", (format("%1%") % v.integer).str()));
+            break;
+
+        case tBool:
+            doc.writeEmptyElement("bool", singletonAttrs("value", v.boolean ? "true" : "false"));
+            break;
+
+        case tString:
+            /* !!! show the context? */
+            copyContext(v, context);
+            doc.writeEmptyElement("string", singletonAttrs("value", v.string.s));
+            break;
+
+        case tPath:
+            doc.writeEmptyElement("path", singletonAttrs("value", v.path));
+            break;
+
+        case tNull:
+            doc.writeEmptyElement("null");
+            break;
+
+        case tAttrs:
+            if (state.isDerivation(v)) {
+                XMLAttrs xmlAttrs;
+            
+                Bindings::iterator a = v.attrs->find(state.symbols.create("derivation"));
+
+                Path drvPath;
+                a = v.attrs->find(state.sDrvPath);
+                if (a != v.attrs->end()) {
+                    if (strict) state.forceValue(*a->value);
+                    if (a->value->type == tString)
+                        xmlAttrs["drvPath"] = drvPath = a->value->string.s;
+                }
+        
+                a = v.attrs->find(state.sOutPath);
+                if (a != v.attrs->end()) {
+                    if (strict) state.forceValue(*a->value);
+                    if (a->value->type == tString)
+                        xmlAttrs["outPath"] = a->value->string.s;
+                }
+
+                XMLOpenElement _(doc, "derivation", xmlAttrs);
+
+                if (drvPath != "" && drvsSeen.find(drvPath) == drvsSeen.end()) {
+                    drvsSeen.insert(drvPath);
+                    showAttrs(state, strict, location, *v.attrs, doc, context, drvsSeen);
+                } else
+                    doc.writeEmptyElement("repeated");
+            }
+
+            else {
+                XMLOpenElement _(doc, "attrs");
+                showAttrs(state, strict, location, *v.attrs, doc, context, drvsSeen);
+            }
+            
+            break;
+
+        case tList: {
+            XMLOpenElement _(doc, "list");
+            for (unsigned int n = 0; n < v.list.length; ++n)
+                printValueAsXML(state, strict, location, *v.list.elems[n], doc, context, drvsSeen);
+            break;
+        }
+
+        case tLambda: {
+            XMLAttrs xmlAttrs;
+            if (location) posToXML(xmlAttrs, v.lambda.fun->pos);
+            XMLOpenElement _(doc, "function", xmlAttrs);
+            
+            if (v.lambda.fun->matchAttrs) {
+                XMLAttrs attrs;
+                if (!v.lambda.fun->arg.empty()) attrs["name"] = v.lambda.fun->arg;
+                if (v.lambda.fun->formals->ellipsis) attrs["ellipsis"] = "1";
+                XMLOpenElement _(doc, "attrspat", attrs);
+                foreach (Formals::Formals_::iterator, i, v.lambda.fun->formals->formals)
+                    doc.writeEmptyElement("attr", singletonAttrs("name", i->name));
+            } else
+                doc.writeEmptyElement("varpat", singletonAttrs("name", v.lambda.fun->arg));
+            
+            break;
+        }
+
+        default:
+            doc.writeEmptyElement("unevaluated");
+    }
+}
+
+
+void printValueAsXML(EvalState & state, bool strict, bool location,
+    Value & v, std::ostream & out, PathSet & context)
+{
+    XMLWriter doc(true, out);
+    XMLOpenElement root(doc, "expr");
+    PathSet drvsSeen;    
+    printValueAsXML(state, strict, location, v, doc, context, drvsSeen);
+}
+
+ 
+}

src/libexpr/value-to-xml.hh

@@ -0,0 +1,17 @@
+#ifndef __VALUE_TO_XML_H
+#define __VALUE_TO_XML_H
+
+#include "nixexpr.hh"
+#include "eval.hh"
+
+#include <string>
+#include <map>
+
+namespace nix {
+
+void printValueAsXML(EvalState & state, bool strict, bool location,
+    Value & v, std::ostream & out, PathSet & context);
+    
+}
+
+#endif /* !__VALUE_TO_XML_H */

src/libmain/Makefile.am

@@ -2,7 +2,7 @@ pkglib_LTLIBRARIES = libmain.la
 
 libmain_la_SOURCES = shared.cc
 
-libmain_la_LIBADD = ../libstore/libstore.la
+libmain_la_LIBADD = ../libstore/libstore.la @boehmgc_lib@
 
 pkginclude_HEADERS = shared.hh
 
@@ -15,5 +15,5 @@ AM_CXXFLAGS = \
  -DNIX_LIBEXEC_DIR=\"$(libexecdir)\" \
  -DNIX_BIN_DIR=\"$(bindir)\" \
  -DNIX_VERSION=\"$(VERSION)\" \
- -I$(srcdir)/.. ${aterm_include} -I$(srcdir)/../libutil \
+ -I$(srcdir)/.. -I$(srcdir)/../libutil \
  -I$(srcdir)/../libstore

src/libmain/shared.cc

@@ -13,7 +13,9 @@
 #include <sys/stat.h>
 #include <unistd.h>
 
-#include <aterm2.h>
+#if HAVE_BOEHMGC
+#include <gc/gc.h>
+#endif
 
 
 namespace nix {
@@ -52,25 +54,26 @@ void printGCWarning()
 
 void printMissing(const PathSet & paths)
 {
-    unsigned long long downloadSize;
+    unsigned long long downloadSize, narSize;
     PathSet willBuild, willSubstitute, unknown;
-    queryMissing(paths, willBuild, willSubstitute, unknown, downloadSize);
+    queryMissing(paths, willBuild, willSubstitute, unknown, downloadSize, narSize);
 
     if (!willBuild.empty()) {
-        printMsg(lvlInfo, format("the following derivations will be built:"));
+        printMsg(lvlInfo, format("these derivations will be built:"));
         foreach (PathSet::iterator, i, willBuild)
             printMsg(lvlInfo, format("  %1%") % *i);
     }
 
     if (!willSubstitute.empty()) {
-        printMsg(lvlInfo, format("the following paths will be downloaded/copied (%.2f MiB):") %
-            (downloadSize / (1024.0 * 1024.0)));
+        printMsg(lvlInfo, format("these paths will be downloaded/copied (%.2f MiB download, %.2f MiB unpacked):")
+            % (downloadSize / (1024.0 * 1024.0))
+            % (narSize / (1024.0 * 1024.0)));
         foreach (PathSet::iterator, i, willSubstitute)
             printMsg(lvlInfo, format("  %1%") % *i);
     }
 
     if (!unknown.empty()) {
-        printMsg(lvlInfo, format("don't know how to build the following paths%1%:")
+        printMsg(lvlInfo, format("don't know how to build these paths%1%:")
             % (readOnlyMode ? " (may be caused by read-only store access)" : ""));
         foreach (PathSet::iterator, i, unknown)
             printMsg(lvlInfo, format("  %1%") % *i);
@@ -87,9 +90,6 @@ static void setLogType(string lt)
 }
 
 
-void initDerivationsHelpers();
-
-
 static void closeStore()
 {
     try {
@@ -140,6 +140,7 @@ static void initAndRun(int argc, char * * argv)
     /* Get some settings from the configuration file. */
     thisSystem = querySetting("system", SYSTEM);
     maxBuildJobs = queryIntSetting("build-max-jobs", 1);
+    buildCores = queryIntSetting("build-cores", 1);
     maxSilentTime = queryIntSetting("build-max-silent-time", 0);
 
     /* Catch SIGINT. */
@@ -176,9 +177,6 @@ static void initAndRun(int argc, char * * argv)
     string lt = getEnv("NIX_LOG_TYPE");
     if (lt != "") setLogType(lt);
 
-    /* ATerm stuff.  !!! find a better place to put this */
-    initDerivationsHelpers();
-    
     /* Put the arguments in a vector. */
     Strings args, remaining;
     while (argc--) args.push_back(*argv++);
@@ -203,17 +201,16 @@ static void initAndRun(int argc, char * * argv)
     remaining.clear();
 
     /* Process default options. */
+    int verbosityDelta = 0;
     for (Strings::iterator i = args.begin(); i != args.end(); ++i) {
         string arg = *i;
-        if (arg == "--verbose" || arg == "-v")
-            verbosity = (Verbosity) ((int) verbosity + 1);
+        if (arg == "--verbose" || arg == "-v") verbosityDelta++;
+        else if (arg == "--quiet") verbosityDelta--;
         else if (arg == "--log-type") {
             ++i;
             if (i == args.end()) throw UsageError("`--log-type' requires an argument");
             setLogType(*i);
         }
-        else if (arg == "--build-output" || arg == "-B")
-            ; /* !!! obsolete - remove eventually */
         else if (arg == "--no-build-output" || arg == "-Q")
             buildVerbosity = lvlVomit;
         else if (arg == "--print-build-trace")
@@ -234,6 +231,8 @@ static void initAndRun(int argc, char * * argv)
             tryFallback = true;
         else if (arg == "--max-jobs" || arg == "-j")
             maxBuildJobs = getIntArg<unsigned int>(arg, i, args.end());
+        else if (arg == "--cores")
+            buildCores = getIntArg<unsigned int>(arg, i, args.end());
         else if (arg == "--readonly-mode")
             readOnlyMode = true;
         else if (arg == "--max-silent-time")
@@ -252,6 +251,9 @@ static void initAndRun(int argc, char * * argv)
         else remaining.push_back(arg);
     }
 
+    verbosityDelta += queryIntSetting("verbosity", lvlInfo);
+    verbosity = (Verbosity) (verbosityDelta < 0 ? 0 : verbosityDelta);
+    
     /* Automatically clean up the temporary roots file when we
        exit. */
     RemoveTempRoots removeTempRoots __attribute__((unused));
@@ -319,6 +321,14 @@ static void setuidInit()
 }
 
 
+/* Called when the Boehm GC runs out of memory. */
+static void * oomHandler(size_t requested)
+{
+    /* Convert this to a proper C++ exception. */
+    throw std::bad_alloc();
+}
+
+
 }
 
 
@@ -333,10 +343,6 @@ int main(int argc, char * * argv)
     if (argc == 0) abort();
     setuidInit();
     
-    /* ATerm setup. */
-    ATerm bottomOfStack;
-    ATinit(argc, argv, &bottomOfStack);
-
     /* Turn on buffering for cerr. */
 #if HAVE_PUBSETBUF
     std::cerr.rdbuf()->pubsetbuf(buf, sizeof(buf));
@@ -344,6 +350,26 @@ int main(int argc, char * * argv)
 
     std::ios::sync_with_stdio(false);
 
+#if HAVE_BOEHMGC
+    /* Initialise the Boehm garbage collector.  This isn't necessary
+       on most platforms, but for portability we do it anyway. */
+    GC_INIT();
+
+    GC_oom_fn = oomHandler;
+
+    /* Set the initial heap size to something fairly big (384 MiB) so
+       that in most cases we don't need to garbage collect at all.
+       (Collection has a fairly significant overhead, some.)  The heap
+       size can be overriden through libgc's GC_INITIAL_HEAP_SIZE
+       environment variable.  We should probably also provide a
+       nix.conf setting for this.  Note that GC_expand_hp() causes a
+       lot of virtual, but not physical (resident) memory to be
+       allocated.  This might be a problem on systems that don't
+       overcommit. */
+    if (!getenv("GC_INITIAL_HEAP_SIZE"))
+        GC_expand_hp(384 * 1024 * 1024);
+#endif
+
     try {
         try {
             initAndRun(argc, argv);
@@ -367,7 +393,7 @@ int main(int argc, char * * argv)
         printMsg(lvlError, format("error: %1%%2%") % (showTrace ? e.prefix() : "") % e.msg());
         if (e.prefix() != "" && !showTrace)
             printMsg(lvlError, "(use `--show-trace' to show detailed location information)");
-        return 1;
+        return e.status;
     } catch (std::exception & e) {
         printMsg(lvlError, format("error: %1%") % e.what());
         return 1;

src/libmain/shared.hh

@@ -1,7 +1,7 @@
 #ifndef __SHARED_H
 #define __SHARED_H
 
-#include "types.hh"
+#include "util.hh"
 
 #include <signal.h>
 

src/libstore/Makefile.am

@@ -10,14 +10,14 @@ pkginclude_HEADERS = \
   globals.hh references.hh pathlocks.hh \
   worker-protocol.hh
 
-libstore_la_LIBADD = ../libutil/libutil.la ../boost/format/libformat.la @ADDITIONAL_NETWORK_LIBS@
+libstore_la_LIBADD = ../libutil/libutil.la ../boost/format/libformat.la ${aterm_lib} ${sqlite_lib}
 
-BUILT_SOURCES = derivations-ast.cc derivations-ast.hh
-
-EXTRA_DIST = derivations-ast.def derivations-ast.cc
+EXTRA_DIST = schema.sql
 
 AM_CXXFLAGS = -Wall \
- -I$(srcdir)/.. ${aterm_include} -I$(srcdir)/../libutil
+ ${sqlite_include} -I$(srcdir)/.. -I$(srcdir)/../libutil
 
-derivations-ast.cc derivations-ast.hh: ../aterm-helper.pl derivations-ast.def
-	$(perl) $(srcdir)/../aterm-helper.pl derivations-ast.hh derivations-ast.cc < $(srcdir)/derivations-ast.def
+local-store.lo: schema.sql.hh
+
+%.sql.hh: %.sql
+	../bin2c/bin2c schema < $< > $@ || (rm $@ && exit 1)

src/libstore/build.cc

@@ -25,6 +25,7 @@
 #include <unistd.h>
 #include <errno.h>
 #include <stdio.h>
+#include <cstring>
 
 #include <pwd.h>
 #include <grp.h>
@@ -63,6 +64,7 @@ static const uid_t rootUserId = 0;
 
 /* Forward definition. */
 class Worker;
+class HookInstance;
 
 
 /* A pointer to a goal. */
@@ -212,8 +214,14 @@ public:
 
     bool cacheFailure;
 
+    /* Set if at least one derivation had a BuildError (i.e. permanent
+       failure). */
+    bool permanentFailure;
+
     LocalStore & store;
 
+    boost::shared_ptr<HookInstance> hook;
+    
     Worker(LocalStore & store);
     ~Worker();
 
@@ -262,12 +270,13 @@ public:
 
     /* Wait for input to become available. */
     void waitForInput();
-    
+
+    unsigned int exitStatus();
 };
 
 
 MakeError(SubstError, Error)
-MakeError(BuildError, Error) /* denoted a permanent build failure */
+MakeError(BuildError, Error) /* denotes a permanent build failure */
 
 
 //////////////////////////////////////////////////////////////////////
@@ -614,6 +623,107 @@ void deletePathWrapped(const Path & path)
 //////////////////////////////////////////////////////////////////////
 
 
+struct HookInstance
+{
+    /* Pipes for talking to the build hook. */
+    Pipe toHook;
+
+    /* Pipe for the hook's standard output/error. */
+    Pipe fromHook;
+
+    /* Pipe for the builder's standard output/error. */
+    Pipe builderOut;
+    
+    /* The process ID of the hook. */
+    Pid pid;
+
+    HookInstance();
+
+    ~HookInstance();
+};
+
+
+HookInstance::HookInstance()
+{
+    debug("starting build hook");
+    
+    Path buildHook = absPath(getEnv("NIX_BUILD_HOOK"));
+    
+    /* Create a pipe to get the output of the child. */
+    fromHook.create();
+    
+    /* Create the communication pipes. */
+    toHook.create();
+
+    /* Create a pipe to get the output of the builder. */
+    builderOut.create();
+
+    /* Fork the hook. */
+    pid = fork();
+    switch (pid) {
+        
+    case -1:
+        throw SysError("unable to fork");
+
+    case 0:
+        try { /* child */
+
+            commonChildInit(fromHook);
+
+            if (chdir("/") == -1) throw SysError("changing into `/");
+            
+            /* Dup the communication pipes. */
+            toHook.writeSide.close();
+            if (dup2(toHook.readSide, STDIN_FILENO) == -1)
+                throw SysError("dupping to-hook read side");
+
+            /* Use fd 4 for the builder's stdout/stderr. */
+            builderOut.readSide.close();
+            if (dup2(builderOut.writeSide, 4) == -1)
+                throw SysError("dupping builder's stdout/stderr");
+            
+            execl(buildHook.c_str(), buildHook.c_str(), thisSystem.c_str(),
+                (format("%1%") % maxSilentTime).str().c_str(),
+                (format("%1%") % printBuildTrace).str().c_str(),
+                NULL);
+            
+            throw SysError(format("executing `%1%'") % buildHook);
+            
+        } catch (std::exception & e) {
+            std::cerr << format("build hook error: %1%") % e.what() << std::endl;
+        }
+        quickExit(1);
+    }
+    
+    /* parent */
+    pid.setSeparatePG(true);
+    pid.setKillSignal(SIGTERM);
+    fromHook.writeSide.close();
+    toHook.readSide.close();
+}
+
+
+HookInstance::~HookInstance()
+{
+    try {
+        /* Cleanly shut down the hook by closing its stdin if it's not
+           already building.  Otherwise pid's destructor will kill
+           it. */
+        if (pid != -1 && toHook.writeSide != -1) {
+            toHook.writeSide.close();
+            pid.wait(true);
+        }
+    } catch (...) {
+        ignoreException();
+    }
+}
+
+
+//////////////////////////////////////////////////////////////////////
+
+
+typedef enum {rpAccept, rpDecline, rpPostpone} HookReply;
+
 class DerivationGoal : public Goal
 {
 private:
@@ -648,14 +758,11 @@ private:
     AutoCloseFD fdLogFile;
 
     /* Pipe for the builder's standard output/error. */
-    Pipe logPipe;
-
-    /* Whether we're building using a build hook. */
-    bool usingBuildHook;
-
-    /* Pipes for talking to the build hook (if any). */
-    Pipe toHook;
+    Pipe builderOut;
 
+    /* The build hook. */
+    boost::shared_ptr<HookInstance> hook;
+    
     /* Whether we're currently doing a chroot build. */
     bool useChroot;
     
@@ -693,12 +800,8 @@ private:
     void buildDone();
 
     /* Is the build hook willing to perform the build? */
-    typedef enum {rpAccept, rpDecline, rpPostpone} HookReply;
     HookReply tryBuildHook();
 
-    /* Synchronously wait for a build hook to finish. */
-    void terminateBuildHook(bool kill = false);
-
     /* Start building a derivation. */
     void startBuilder();
 
@@ -710,10 +813,6 @@ private:
     /* Open a log file and a pipe to it. */
     Path openLogFile();
 
-    /* Common initialisation to be performed in child processes (i.e.,
-       both in builders and in build hooks). */
-    void initChild();
-    
     /* Delete the temporary directory, if we have one. */
     void deleteTmpDir(bool force);
 
@@ -741,6 +840,7 @@ DerivationGoal::DerivationGoal(const Path & drvPath, Worker & worker)
     trace("created");
 }
 
+
 DerivationGoal::~DerivationGoal()
 {
     /* Careful: we should never ever throw an exception from a
@@ -753,6 +853,7 @@ DerivationGoal::~DerivationGoal()
     }
 }
 
+
 void DerivationGoal::killChild()
 {
     if (pid != -1) {
@@ -777,6 +878,8 @@ void DerivationGoal::killChild()
         
         assert(pid == -1);
     }
+
+    hook.reset();
 }
 
 
@@ -886,7 +989,10 @@ void DerivationGoal::outputsSubstituted()
     foreach (PathSet::iterator, i, drv.inputSrcs)
         addWaitee(worker.makeSubstitutionGoal(*i));
 
-    state = &DerivationGoal::inputsRealised;
+    if (waitees.empty()) /* to prevent hang (no wake-up event) */
+        inputsRealised();
+    else
+        state = &DerivationGoal::inputsRealised;
 }
 
 
@@ -960,6 +1066,16 @@ PathSet outputPaths(const DerivationOutputs & outputs)
 }
 
 
+static bool canBuildLocally(const string & platform)
+{
+    return platform == thisSystem 
+#ifdef CAN_DO_LINUX32_BUILDS
+        || (platform == "i686-linux" && thisSystem == "x86_64-linux")
+#endif
+        ;
+}
+
+
 void DerivationGoal::tryToBuild()
 {
     trace("trying to build");
@@ -1027,28 +1143,36 @@ void DerivationGoal::tryToBuild()
     foreach (DerivationOutputs::iterator, i, drv.outputs)
         if (pathFailed(i->second.path)) return;
 
+    /* Don't do a remote build if the derivation has the attribute
+       `preferLocalBuild' set. */
+    bool preferLocalBuild =
+        drv.env["preferLocalBuild"] == "1" && canBuildLocally(drv.platform);
+
     /* Is the build hook willing to accept this job? */
-    usingBuildHook = true;
-    switch (tryBuildHook()) {
-        case rpAccept:
-            /* Yes, it has started doing so.  Wait until we get EOF
-               from the hook. */
-            state = &DerivationGoal::buildDone;
-            return;
-        case rpPostpone:
-            /* Not now; wait until at least one child finishes. */
-            worker.waitForAWhile(shared_from_this());
-            outputLocks.unlock();
-            return;
-        case rpDecline:
-            /* We should do it ourselves. */
-            break;
+    if (!preferLocalBuild) {
+        switch (tryBuildHook()) {
+            case rpAccept:
+                /* Yes, it has started doing so.  Wait until we get
+                   EOF from the hook. */
+                state = &DerivationGoal::buildDone;
+                return;
+            case rpPostpone:
+                /* Not now; wait until at least one child finishes or
+                   the wake-up timeout expires. */
+                worker.waitForAWhile(shared_from_this());
+                outputLocks.unlock();
+                return;
+            case rpDecline:
+                /* We should do it ourselves. */
+                break;
+        }
     }
-
-    usingBuildHook = false;
-
-    /* Make sure that we are allowed to start a build. */
-    if (worker.getNrLocalBuilds() >= maxBuildJobs) {
+    
+    /* Make sure that we are allowed to start a build.  If this
+       derivation prefers to be done locally, do it even if
+       maxBuildJobs is 0. */
+    unsigned int curBuilds = worker.getNrLocalBuilds();
+    if (curBuilds >= maxBuildJobs && !(preferLocalBuild && curBuilds == 0)) {
         worker.waitForBuildSlot(shared_from_this());
         outputLocks.unlock();
         return;
@@ -1066,6 +1190,7 @@ void DerivationGoal::tryToBuild()
         if (printBuildTrace)
             printMsg(lvlError, format("@ build-failed %1% %2% %3% %4%")
                 % drvPath % drv.outputs["out"].path % 0 % e.msg());
+        worker.permanentFailure = true;
         amDone(ecFailed);
         return;
     }
@@ -1084,18 +1209,29 @@ void DerivationGoal::buildDone()
        to have terminated.  In fact, the builder could also have
        simply have closed its end of the pipe --- just don't do that
        :-) */
-    /* !!! this could block! security problem! solution: kill the
-       child */
-    pid_t savedPid = pid;
-    int status = pid.wait(true);
+    int status;
+    pid_t savedPid;
+    if (hook) {
+        savedPid = hook->pid;
+        status = hook->pid.wait(true);
+    } else {
+        /* !!! this could block! security problem! solution: kill the
+           child */
+        savedPid = pid;
+        status = pid.wait(true);
+    }
 
     debug(format("builder process for `%1%' finished") % drvPath);
 
     /* So the child is gone now. */
     worker.childTerminated(savedPid);
-
+    
     /* Close the read side of the logger pipe. */
-    logPipe.readSide.close();
+    if (hook) {
+        hook->builderOut.readSide.close();
+        hook->fromHook.readSide.close();
+    }
+    else builderOut.readSide.close();
 
     /* Close the log file. */
     fdLogFile.close();
@@ -1168,11 +1304,11 @@ void DerivationGoal::buildDone()
         /* When using a build hook, the hook will return a remote
            build failure using exit code 100.  Anything else is a hook
            problem. */
-        bool hookError = usingBuildHook &&
+        bool hookError = hook &&
             (!WIFEXITED(status) || WEXITSTATUS(status) != 100);
         
         if (printBuildTrace) {
-            if (usingBuildHook && hookError)
+            if (hook && hookError)
                 printMsg(lvlError, format("@ hook-failed %1% %2% %3% %4%")
                     % drvPath % drv.outputs["out"].path % status % e.msg());
             else
@@ -1191,6 +1327,7 @@ void DerivationGoal::buildDone()
             foreach (DerivationOutputs::iterator, i, drv.outputs)
                 worker.store.registerFailedPath(i->second.path);
         
+        worker.permanentFailure = !hookError && !fixedOutput;
         amDone(ecFailed);
         return;
     }
@@ -1207,162 +1344,85 @@ void DerivationGoal::buildDone()
 }
 
 
-DerivationGoal::HookReply DerivationGoal::tryBuildHook()
+HookReply DerivationGoal::tryBuildHook()
 {
-    if (!useBuildHook) return rpDecline;
-    Path buildHook = getEnv("NIX_BUILD_HOOK");
-    if (buildHook == "") return rpDecline;
-    buildHook = absPath(buildHook);
+    if (!useBuildHook || getEnv("NIX_BUILD_HOOK") == "") return rpDecline;
 
-    /* Create a directory where we will store files used for
-       communication between us and the build hook. */
-    tmpDir = createTempDir();
-    
-    /* Create the log file and pipe. */
-    Path logFile = openLogFile();
+    if (!worker.hook)
+        worker.hook = boost::shared_ptr<HookInstance>(new HookInstance);
 
-    /* Create the communication pipes. */
-    toHook.create();
+    /* Tell the hook about system features (beyond the system type)
+       required from the build machine.  (The hook could parse the
+       drv file itself, but this is easier.) */
+    Strings features = tokenizeString(drv.env["requiredSystemFeatures"]);
+    foreach (Strings::iterator, i, features) checkStoreName(*i); /* !!! abuse */
 
-    /* Fork the hook. */
-    pid = fork();
-    switch (pid) {
-        
-    case -1:
-        throw SysError("unable to fork");
-
-    case 0:
-        try { /* child */
-
-            initChild();
-
-            string s;
-            foreach (DerivationOutputs::const_iterator, i, drv.outputs)
-                s += i->second.path + " ";
-            if (setenv("NIX_HELD_LOCKS", s.c_str(), 1))
-                throw SysError("setting an environment variable");
-
-            execl(buildHook.c_str(), buildHook.c_str(),
-                (worker.getNrLocalBuilds() < maxBuildJobs ? (string) "1" : "0").c_str(),
-                thisSystem.c_str(),
-                drv.platform.c_str(),
-                drvPath.c_str(),
-                (format("%1%") % maxSilentTime).str().c_str(),
-                NULL);
-            
-            throw SysError(format("executing `%1%'") % buildHook);
-            
-        } catch (std::exception & e) {
-            std::cerr << format("build hook error: %1%") % e.what() << std::endl;
-        }
-        quickExit(1);
-    }
-    
-    /* parent */
-    pid.setSeparatePG(true);
-    pid.setKillSignal(SIGTERM);
-    logPipe.writeSide.close();
-    worker.childStarted(shared_from_this(),
-        pid, singleton<set<int> >(logPipe.readSide), false, false);
-
-    toHook.readSide.close();
+    /* Send the request to the hook. */
+    writeLine(worker.hook->toHook.writeSide, (format("%1% %2% %3% %4%")
+        % (worker.getNrLocalBuilds() < maxBuildJobs ? "1" : "0")
+        % drv.platform % drvPath % concatStringsSep(",", features)).str());
 
     /* Read the first line of input, which should be a word indicating
        whether the hook wishes to perform the build. */
     string reply;
-    try {
-        while (true) {
-            string s = readLine(logPipe.readSide);
-            if (string(s, 0, 2) == "# ") {
-                reply = string(s, 2);
-                break;
-            }
-            handleChildOutput(logPipe.readSide, s + "\n");
+    while (true) {
+        string s = readLine(worker.hook->fromHook.readSide);
+        if (string(s, 0, 2) == "# ") {
+            reply = string(s, 2);
+            break;
         }
-    } catch (Error & e) {
-        terminateBuildHook(true);
-        throw;
+        s += "\n";
+        writeToStderr((unsigned char *) s.c_str(), s.size());
     }
 
     debug(format("hook reply is `%1%'") % reply);
 
-    if (reply == "decline" || reply == "postpone") {
-        /* Clean up the child.  !!! hacky / should verify */
-        terminateBuildHook();
+    if (reply == "decline" || reply == "postpone")
         return reply == "decline" ? rpDecline : rpPostpone;
-    }
+    else if (reply != "accept")
+        throw Error(format("bad hook reply `%1%'") % reply);
 
-    else if (reply == "accept") {
+    printMsg(lvlTalkative, format("using hook to build path(s) %1%")
+        % showPaths(outputPaths(drv.outputs)));
 
-        printMsg(lvlInfo, format("using hook to build path(s) %1%")
-            % showPaths(outputPaths(drv.outputs)));
+    hook = worker.hook;
+    worker.hook.reset();
         
-        /* Write the information that the hook needs to perform the
-           build, i.e., the set of input paths, the set of output
-           paths, and the references (pointer graph) in the input
-           paths. */
+    /* Tell the hook all the inputs that have to be copied to the
+       remote system.  This unfortunately has to contain the entire
+       derivation closure to ensure that the validity invariant holds
+       on the remote system.  (I.e., it's unfortunate that we have to
+       list it since the remote system *probably* already has it.) */
+    PathSet allInputs;
+    allInputs.insert(inputPaths.begin(), inputPaths.end());
+    computeFSClosure(drvPath, allInputs);
         
-        Path inputListFN = tmpDir + "/inputs";
-        Path outputListFN = tmpDir + "/outputs";
-        Path referencesFN = tmpDir + "/references";
-
-        /* The `inputs' file lists all inputs that have to be copied
-           to the remote system.  This unfortunately has to contain
-           the entire derivation closure to ensure that the validity
-           invariant holds on the remote system.  (I.e., it's
-           unfortunate that we have to list it since the remote system
-           *probably* already has it.) */
-        PathSet allInputs;
-        allInputs.insert(inputPaths.begin(), inputPaths.end());
-        computeFSClosure(drvPath, allInputs);
+    string s;
+    foreach (PathSet::iterator, i, allInputs) s += *i + " ";
+    writeLine(hook->toHook.writeSide, s);
         
-        string s;
-        foreach (PathSet::iterator, i, allInputs) s += *i + "\n";
+    /* Tell the hooks the outputs that have to be copied back from the
+       remote system. */
+    s = "";
+    foreach (DerivationOutputs::iterator, i, drv.outputs)
+        s += i->second.path + " ";
+    writeLine(hook->toHook.writeSide, s);
+    
+    hook->toHook.writeSide.close();
+
+    /* Create the log file and pipe. */
+    Path logFile = openLogFile();
+
+    set<int> fds;
+    fds.insert(hook->fromHook.readSide);
+    fds.insert(hook->builderOut.readSide);
+    worker.childStarted(shared_from_this(), hook->pid, fds, false, false);
+    
+    if (printBuildTrace)
+        printMsg(lvlError, format("@ build-started %1% %2% %3% %4%")
+            % drvPath % drv.outputs["out"].path % drv.platform % logFile);
         
-        writeFile(inputListFN, s);
-
-        /* The `outputs' file lists all outputs that have to be copied
-           from the remote system. */
-        s = "";
-        foreach (DerivationOutputs::iterator, i, drv.outputs)
-            s += i->second.path + "\n";
-        writeFile(outputListFN, s);
-
-        /* The `references' file has exactly the format accepted by
-           `nix-store --register-validity'. */
-        writeFile(referencesFN,
-            makeValidityRegistration(allInputs, true, false));
-
-        /* Tell the hook to proceed. */
-        writeLine(toHook.writeSide, "okay");
-        toHook.writeSide.close();
-
-        if (printBuildTrace)
-            printMsg(lvlError, format("@ build-started %1% %2% %3% %4%")
-                % drvPath % drv.outputs["out"].path % drv.platform % logFile);
-        
-        return rpAccept;
-    }
-
-    else throw Error(format("bad hook reply `%1%'") % reply);
-}
-
-
-void DerivationGoal::terminateBuildHook(bool kill)
-{
-    debug("terminating build hook");
-    pid_t savedPid = pid;
-    if (kill)
-        pid.kill();
-    else
-        pid.wait(true);
-    /* `false' means don't wake up waiting goals, since we want to
-       keep this build slot ourselves. */
-    worker.childTerminated(savedPid, false);
-    toHook.writeSide.close();
-    fdLogFile.close();
-    logPipe.readSide.close();
-    deleteTmpDir(true); /* get rid of the hook's temporary directory */
+    return rpAccept;    
 }
 
 
@@ -1379,11 +1439,7 @@ void DerivationGoal::startBuilder()
         format("building path(s) %1%") % showPaths(outputPaths(drv.outputs)))
     
     /* Right platform? */
-    if (drv.platform != thisSystem 
-#ifdef CAN_DO_LINUX32_BUILDS
-        && !(drv.platform == "i686-linux" && thisSystem == "x86_64-linux")
-#endif
-        )
+    if (!canBuildLocally(drv.platform))
         throw Error(
             format("a `%1%' is required to build `%3%', but I am a `%2%'")
             % drv.platform % thisSystem % drvPath);
@@ -1411,6 +1467,9 @@ void DerivationGoal::startBuilder()
        in the store or in the build directory). */
     env["NIX_STORE"] = nixStore;
 
+    /* The maximum number of cores to utilize for parallel building. */
+    env["NIX_BUILD_CORES"] = (format("%d") % buildCores).str();
+
     /* Add all bindings specified in the derivation. */
     foreach (StringPairs::iterator, i, drv.env)
         env[i->first] = i->second;
@@ -1495,7 +1554,7 @@ void DerivationGoal::startBuilder()
 
         /* Write closure info to `fileName'. */
         writeFile(tmpDir + "/" + fileName,
-            makeValidityRegistration(paths, false, false));
+            worker.store.makeValidityRegistration(paths, false, false));
     }
 
     
@@ -1545,6 +1604,9 @@ void DerivationGoal::startBuilder()
 
     if (fixedOutput) useChroot = false;
 
+    /* Hack to allow derivations to disable chroot builds. */
+    if (drv.env["__noChroot"] == "1") useChroot = false;
+
     if (useChroot) {
 #if CHROOT_ENABLED
         /* Create a temporary directory in which we set up the chroot
@@ -1566,16 +1628,23 @@ void DerivationGoal::startBuilder()
         createDirs(chrootTmpDir);
         chmod(chrootTmpDir, 01777);
 
-        /* Create a /etc/passwd with entries for the build user and
-           the nobody account.  The latter is kind of a hack to
-           support Samba-in-QEMU. */
+        /* Create a /etc/passwd with entries for the build user and the
+           nobody account.  The latter is kind of a hack to support
+           Samba-in-QEMU. */
         createDirs(chrootRootDir + "/etc");
 
         writeFile(chrootRootDir + "/etc/passwd",
             (format(
-                "nixbld:x:%1%:65534:Nix build user:/:/noshell\n"
+                "nixbld:x:%1%:%2%:Nix build user:/:/noshell\n"
                 "nobody:x:65534:65534:Nobody:/:/noshell\n")
-                % (buildUser.enabled() ? buildUser.getUID() : getuid())).str());
+                % (buildUser.enabled() ? buildUser.getUID() : getuid())
+                % (buildUser.enabled() ? buildUser.getGID() : getgid())).str());
+
+	/* Declare the build user's group so that programs get a consistent
+	   view of the system (e.g., "id -gn"). */
+        writeFile(chrootRootDir + "/etc/group",
+            (format("nixbld:!:%1%:\n")
+                % (buildUser.enabled() ? buildUser.getGID() : getgid())).str());
 
         /* Bind-mount a user-configurable set of directories from the
            host file system.  The `/dev/pts' directory must be mounted
@@ -1634,9 +1703,12 @@ void DerivationGoal::startBuilder()
     printMsg(lvlChatty, format("executing builder `%1%'") %
         drv.builder);
 
-    /* Create the log file and pipe. */
+    /* Create the log file. */
     Path logFile = openLogFile();
     
+    /* Create a pipe to get the output of the builder. */
+    builderOut.create();
+
     /* Fork a child to build the package.  Note that while we
        currently use forks to run and wait for the children, it
        shouldn't be hard to use threads for this on systems where
@@ -1650,7 +1722,7 @@ void DerivationGoal::startBuilder()
     case 0:
 
         /* Warning: in the child we should absolutely not make any
-           Berkeley DB calls! */
+           SQLite calls! */
 
         try { /* child */
 
@@ -1677,18 +1749,23 @@ void DerivationGoal::startBuilder()
                         throw SysError(format("bind mount from `%1%' to `%2%' failed") % source % target);
                 }
                     
-                /* Do the chroot().  initChild() will do a chdir() to
-                   the temporary build directory to make sure the
-                   current directory is in the chroot.  (Actually the
-                   order doesn't matter, since due to the bind mount
-                   tmpDir and tmpRootDit/tmpDir are the same
-                   directories.) */
+                /* Do the chroot().  Below we do a chdir() to the
+                   temporary build directory to make sure the current
+                   directory is in the chroot.  (Actually the order
+                   doesn't matter, since due to the bind mount tmpDir
+                   and tmpRootDit/tmpDir are the same directories.) */
                 if (chroot(chrootRootDir.c_str()) == -1)
                     throw SysError(format("cannot change root directory to `%1%'") % chrootRootDir);
             }
 #endif
             
-            initChild();
+            commonChildInit(builderOut);
+    
+            if (chdir(tmpDir.c_str()) == -1)
+                throw SysError(format("changing into `%1%'") % tmpDir);
+
+            /* Close all other file descriptors. */
+            closeMostFDs(set<int>());
 
 #ifdef CAN_DO_LINUX32_BUILDS
             if (drv.platform == "i686-linux" && thisSystem == "x86_64-linux") {
@@ -1709,10 +1786,10 @@ void DerivationGoal::startBuilder()
             
             /* If we are running in `build-users' mode, then switch to
                the user we allocated above.  Make sure that we drop
-               all root privileges.  Note that initChild() above has
-               closed all file descriptors except std*, so that's
-               safe.  Also note that setuid() when run as root sets
-               the real, effective and saved UIDs. */
+               all root privileges.  Note that above we have closed
+               all file descriptors except std*, so that's safe.  Also
+               note that setuid() when run as root sets the real,
+               effective and saved UIDs. */
             if (buildUser.enabled()) {
                 printMsg(lvlChatty, format("switching to user `%1%'") % buildUser.getUser());
 
@@ -1766,9 +1843,9 @@ void DerivationGoal::startBuilder()
     
     /* parent */
     pid.setSeparatePG(true);
-    logPipe.writeSide.close();
+    builderOut.writeSide.close();
     worker.childStarted(shared_from_this(), pid,
-        singleton<set<int> >(logPipe.readSide), true, true);
+        singleton<set<int> >(builderOut.readSide), true, true);
 
     if (printBuildTrace) {
         printMsg(lvlError, format("@ build-started %1% %2% %3% %4%")
@@ -1800,12 +1877,12 @@ PathSet parseReferenceSpecifiers(const Derivation & drv, string attr)
 void DerivationGoal::computeClosure()
 {
     map<Path, PathSet> allReferences;
-    map<Path, Hash> contentHashes;
+    map<Path, HashResult> contentHashes;
 
     /* When using a build hook, the build hook can register the output
        as valid (by doing `nix-store --import').  If so we don't have
        to do anything here. */
-    if (usingBuildHook) {
+    if (hook) {
         bool allValid = true;
         foreach (DerivationOutputs::iterator, i, drv.outputs)
             if (!worker.store.isValidPath(i->second.path)) allValid = false;
@@ -1857,7 +1934,7 @@ void DerivationGoal::computeClosure()
             if (ht == htUnknown)
                 throw BuildError(format("unknown hash algorithm `%1%'") % algo);
             Hash h = parseHash(ht, i->second.hash);
-            Hash h2 = recursive ? hashPath(ht, path) : hashFile(ht, path);
+            Hash h2 = recursive ? hashPath(ht, path).first : hashFile(ht, path);
             if (h != h2)
                 throw BuildError(
                     format("output path `%1%' should have %2% hash `%3%', instead has `%4%'")
@@ -1871,7 +1948,7 @@ void DerivationGoal::computeClosure()
 	   contained in it.  Compute the SHA-256 NAR hash at the same
 	   time.  The hash is stored in the database so that we can
 	   verify later on whether nobody has messed with the store. */
-        Hash hash;
+        HashResult hash;
         PathSet references = scanForReferences(path, allPaths, hash);
         contentHashes[path] = hash;
 
@@ -1900,14 +1977,18 @@ void DerivationGoal::computeClosure()
     }
 
     /* Register each output path as valid, and register the sets of
-       paths referenced by each of them.  !!! this should be
-       atomic so that either all paths are registered as valid, or
-       none are. */
-    foreach (DerivationOutputs::iterator, i, drv.outputs)
-        worker.store.registerValidPath(i->second.path,
-            contentHashes[i->second.path],
-            allReferences[i->second.path],
-            drvPath);
+       paths referenced by each of them. */
+    ValidPathInfos infos;
+    foreach (DerivationOutputs::iterator, i, drv.outputs) {
+        ValidPathInfo info;
+        info.path = i->second.path;
+        info.hash = contentHashes[i->second.path].first;
+        info.narSize = contentHashes[i->second.path].second;
+        info.references = allReferences[i->second.path];
+        info.deriver = drvPath;
+        infos.push_back(info);
+    }
+    worker.store.registerValidPaths(infos);
 
     /* It is now safe to delete the lock files, since all future
        lockers will see that the output paths are valid; they will not
@@ -1933,32 +2014,10 @@ Path DerivationGoal::openLogFile()
     if (fdLogFile == -1)
         throw SysError(format("creating log file `%1%'") % logFileName);
 
-    /* Create a pipe to get the output of the child. */
-    logPipe.create();
-
     return logFileName;
 }
 
 
-void DerivationGoal::initChild()
-{
-    commonChildInit(logPipe);
-    
-    if (chdir(tmpDir.c_str()) == -1)
-        throw SysError(format("changing into `%1%'") % tmpDir);
-
-    /* When running a hook, dup the communication pipes. */
-    if (usingBuildHook) {
-        toHook.writeSide.close();
-        if (dup2(toHook.readSide, STDIN_FILENO) == -1)
-            throw SysError("dupping to-hook read side");
-    }
-
-    /* Close all other file descriptors. */
-    closeMostFDs(set<int>());
-}
-
-
 void DerivationGoal::deleteTmpDir(bool force)
 {
     if (tmpDir != "") {
@@ -1978,19 +2037,22 @@ void DerivationGoal::deleteTmpDir(bool force)
 
 void DerivationGoal::handleChildOutput(int fd, const string & data)
 {
-    if (fd == logPipe.readSide) {
+    if ((hook && fd == hook->builderOut.readSide) ||
+        (!hook && fd == builderOut.readSide))
+    {
         if (verbosity >= buildVerbosity)
             writeToStderr((unsigned char *) data.c_str(), data.size());
         writeFull(fdLogFile, (unsigned char *) data.c_str(), data.size());
     }
 
-    else abort();
+    if (hook && fd == hook->fromHook.readSide)
+        writeToStderr((unsigned char *) data.c_str(), data.size());
 }
 
 
 void DerivationGoal::handleEOF(int fd)
 {
-    if (fd == logPipe.readSide) worker.wakeUp(shared_from_this());
+    worker.wakeUp(shared_from_this());
 }
 
 
@@ -2334,10 +2396,15 @@ void SubstitutionGoal::finished()
 
     canonicalisePathMetaData(storePath);
 
-    Hash contentHash = hashPath(htSHA256, storePath);
-
-    worker.store.registerValidPath(storePath, contentHash,
-        info.references, info.deriver);
+    HashResult hash = hashPath(htSHA256, storePath);
+    
+    ValidPathInfo info2;
+    info2.path = storePath;
+    info2.hash = hash.first;
+    info2.narSize = hash.second;
+    info2.references = info.references;
+    info2.deriver = info.deriver;
+    worker.store.registerValidPath(info2);
 
     outputLock->setDeletion(true);
     
@@ -2384,6 +2451,7 @@ Worker::Worker(LocalStore & store)
     nrLocalBuilds = 0;
     lastWokenUp = 0;
     cacheFailure = queryBoolSetting("build-cache-failure", false);
+    permanentFailure = false;
 }
 
 
@@ -2628,6 +2696,7 @@ void Worker::waitForInput()
         timeout.tv_sec = std::max((time_t) 0, lastWokenUp + wakeUpInterval - before);
     } else lastWokenUp = 0;
 
+    using namespace std;
     /* Use select() to wait for the input side of any logger pipe to
        become `available'.  Note that `available' (i.e., non-blocking)
        includes EOF. */
@@ -2709,6 +2778,11 @@ void Worker::waitForInput()
 }
 
 
+unsigned int Worker::exitStatus()
+{
+    return permanentFailure ? 100 : 1;
+}
+
 
 //////////////////////////////////////////////////////////////////////
 
@@ -2735,7 +2809,7 @@ void LocalStore::buildDerivations(const PathSet & drvPaths)
         }
             
     if (!failed.empty())
-        throw Error(format("build of %1% failed") % showPaths(failed));
+        throw Error(format("build of %1% failed") % showPaths(failed), worker.exitStatus());
 }
 
 
@@ -2751,7 +2825,7 @@ void LocalStore::ensurePath(const Path & path)
     worker.run(goals);
 
     if (goal->getExitCode() != Goal::ecSuccess)
-        throw Error(format("path `%1%' does not exist and cannot be created") % path);
+        throw Error(format("path `%1%' does not exist and cannot be created") % path, worker.exitStatus());
 }
 
  

src/libstore/derivations-ast.def

@@ -1,10 +0,0 @@
-init initDerivationsHelpers
-
-Derive | ATermList ATermList ATermList string string ATermList ATermList | ATerm |
-
-| string string | ATerm | EnvBinding |
-| string ATermList | ATerm | DerivationInput |
-| string string string string | ATerm | DerivationOutput |
-
-Closure | ATermList ATermList | ATerm | OldClosure |
-| string ATermList | ATerm | OldClosureElem |

src/libstore/derivations.cc

@@ -1,22 +1,12 @@
 #include "derivations.hh"
 #include "store-api.hh"
-#include "aterm.hh"
 #include "globals.hh"
 #include "util.hh"
 
-#include "derivations-ast.hh"
-#include "derivations-ast.cc"
-
 
 namespace nix {
 
 
-Hash hashTerm(ATerm t)
-{
-    return hashString(htSHA256, atPrint(t));
-}
-
-
 Path writeDerivation(const Derivation & drv, const string & name)
 {
     PathSet references;
@@ -27,137 +17,151 @@ Path writeDerivation(const Derivation & drv, const string & name)
        (that can be missing (of course) and should not necessarily be
        held during a garbage collection). */
     string suffix = name + drvExtension;
-    string contents = atPrint(unparseDerivation(drv));
+    string contents = unparseDerivation(drv);
     return readOnlyMode
         ? computeStorePathForText(suffix, contents, references)
         : store->addTextToStore(suffix, contents, references);
 }
 
 
-static void checkPath(const string & s)
+static Path parsePath(std::istream & str)
 {
+    string s = parseString(str);
     if (s.size() == 0 || s[0] != '/')
         throw Error(format("bad path `%1%' in derivation") % s);
+    return s;
 }
     
 
-static void parseStrings(ATermList paths, StringSet & out, bool arePaths)
+static StringSet parseStrings(std::istream & str, bool arePaths)
 {
-    for (ATermIterator i(paths); i; ++i) {
-        if (ATgetType(*i) != AT_APPL)
-            throw badTerm("not a path", *i);
-        string s = aterm2String(*i);
-        if (arePaths) checkPath(s);
-        out.insert(s);
-    }
+    StringSet res;
+    while (!endOfList(str))
+        res.insert(arePaths ? parsePath(str) : parseString(str));
+    return res;
 }
+    
 
-
-/* Shut up warnings. */
-void throwBadDrv(ATerm t) __attribute__ ((noreturn));
-
-void throwBadDrv(ATerm t) 
-{
-    throw badTerm("not a valid derivation", t);
-}
-
-
-Derivation parseDerivation(ATerm t)
+Derivation parseDerivation(const string & s)
 {
     Derivation drv;
-    ATermList outs, inDrvs, inSrcs, args, bnds;
-    ATerm builder, platform;
+    std::istringstream str(s);
+    expect(str, "Derive([");
 
-    if (!matchDerive(t, outs, inDrvs, inSrcs, platform, builder, args, bnds))
-        throwBadDrv(t);
-
-    for (ATermIterator i(outs); i; ++i) {
-        ATerm id, path, hashAlgo, hash;
-        if (!matchDerivationOutput(*i, id, path, hashAlgo, hash))
-            throwBadDrv(t);
+    /* Parse the list of outputs. */
+    while (!endOfList(str)) {
         DerivationOutput out;
-        out.path = aterm2String(path);
-        checkPath(out.path);
-        out.hashAlgo = aterm2String(hashAlgo);
-        out.hash = aterm2String(hash);
-        drv.outputs[aterm2String(id)] = out;
+        expect(str, "("); string id = parseString(str);
+        expect(str, ","); out.path = parsePath(str);
+        expect(str, ","); out.hashAlgo = parseString(str);
+        expect(str, ","); out.hash = parseString(str);
+        expect(str, ")");
+        drv.outputs[id] = out;
     }
 
-    for (ATermIterator i(inDrvs); i; ++i) {
-        ATerm drvPath;
-        ATermList ids;
-        if (!matchDerivationInput(*i, drvPath, ids))
-            throwBadDrv(t);
-        Path drvPath2 = aterm2String(drvPath);
-        checkPath(drvPath2);
-        StringSet ids2;
-        parseStrings(ids, ids2, false);
-        drv.inputDrvs[drvPath2] = ids2;
+    /* Parse the list of input derivations. */
+    expect(str, ",[");
+    while (!endOfList(str)) {
+        expect(str, "(");
+        Path drvPath = parsePath(str);
+        expect(str, ",[");
+        drv.inputDrvs[drvPath] = parseStrings(str, false);
+        expect(str, ")");
+    }
+
+    expect(str, ",["); drv.inputSrcs = parseStrings(str, true);
+    expect(str, ","); drv.platform = parseString(str);
+    expect(str, ","); drv.builder = parseString(str);
+
+    /* Parse the builder arguments. */
+    expect(str, ",[");
+    while (!endOfList(str))
+        drv.args.push_back(parseString(str));
+
+    /* Parse the environment variables. */
+    expect(str, ",[");
+    while (!endOfList(str)) {
+        expect(str, "("); string name = parseString(str);
+        expect(str, ","); string value = parseString(str);
+        expect(str, ")");
+        drv.env[name] = value;
     }
     
-    parseStrings(inSrcs, drv.inputSrcs, true);
-
-    drv.builder = aterm2String(builder);
-    drv.platform = aterm2String(platform);
-    
-    for (ATermIterator i(args); i; ++i) {
-        if (ATgetType(*i) != AT_APPL)
-            throw badTerm("string expected", *i);
-        drv.args.push_back(aterm2String(*i));
-    }
-
-    for (ATermIterator i(bnds); i; ++i) {
-        ATerm s1, s2;
-        if (!matchEnvBinding(*i, s1, s2))
-            throw badTerm("tuple of strings expected", *i);
-        drv.env[aterm2String(s1)] = aterm2String(s2);
-    }
-
+    expect(str, ")");
     return drv;
 }
 
 
-ATerm unparseDerivation(const Derivation & drv)
+static void printString(string & res, const string & s)
 {
-    ATermList outputs = ATempty;
-    for (DerivationOutputs::const_reverse_iterator i = drv.outputs.rbegin();
-         i != drv.outputs.rend(); ++i)
-        outputs = ATinsert(outputs,
-            makeDerivationOutput(
-                toATerm(i->first),
-                toATerm(i->second.path),
-                toATerm(i->second.hashAlgo),
-                toATerm(i->second.hash)));
+    res += '"';
+    for (const char * i = s.c_str(); *i; i++)
+        if (*i == '\"' || *i == '\\') { res += "\\"; res += *i; }
+        else if (*i == '\n') res += "\\n";
+        else if (*i == '\r') res += "\\r";
+        else if (*i == '\t') res += "\\t";
+        else res += *i;
+    res += '"';
+}
 
-    ATermList inDrvs = ATempty;
-    for (DerivationInputs::const_reverse_iterator i = drv.inputDrvs.rbegin();
-         i != drv.inputDrvs.rend(); ++i)
-        inDrvs = ATinsert(inDrvs,
-            makeDerivationInput(
-                toATerm(i->first),
-                toATermList(i->second)));
+
+template<class ForwardIterator>
+static void printStrings(string & res, ForwardIterator i, ForwardIterator j)
+{
+    res += '[';
+    bool first = true;
+    for ( ; i != j; ++i) {
+        if (first) first = false; else res += ',';
+        printString(res, *i);
+    }
+    res += ']';
+}
+
+
+string unparseDerivation(const Derivation & drv)
+{
+    string s;
+    s.reserve(65536);
+    s += "Derive([";
+
+    bool first = true;
+    foreach (DerivationOutputs::const_iterator, i, drv.outputs) {
+        if (first) first = false; else s += ',';
+        s += '('; printString(s, i->first);
+        s += ','; printString(s, i->second.path);
+        s += ','; printString(s, i->second.hashAlgo);
+        s += ','; printString(s, i->second.hash);
+        s += ')';
+    }
+
+    s += "],[";
+    first = true;
+    foreach (DerivationInputs::const_iterator, i, drv.inputDrvs) {
+        if (first) first = false; else s += ',';
+        s += '('; printString(s, i->first);
+        s += ','; printStrings(s, i->second.begin(), i->second.end());
+        s += ')';
+    }
+
+    s += "],";
+    printStrings(s, drv.inputSrcs.begin(), drv.inputSrcs.end());
     
-    ATermList args = ATempty;
-    for (Strings::const_reverse_iterator i = drv.args.rbegin();
-         i != drv.args.rend(); ++i)
-        args = ATinsert(args, toATerm(*i));
+    s += ','; printString(s, drv.platform);
+    s += ','; printString(s, drv.builder);
+    s += ','; printStrings(s, drv.args.begin(), drv.args.end());
 
-    ATermList env = ATempty;
-    for (StringPairs::const_reverse_iterator i = drv.env.rbegin();
-         i != drv.env.rend(); ++i)
-        env = ATinsert(env,
-            makeEnvBinding(
-                toATerm(i->first),
-                toATerm(i->second)));
-
-    return makeDerive(
-        outputs,
-        inDrvs,
-        toATermList(drv.inputSrcs),
-        toATerm(drv.platform),
-        toATerm(drv.builder),
-        args,
-        env);
+    s += ",[";
+    first = true;
+    foreach (StringPairs::const_iterator, i, drv.env) {
+        if (first) first = false; else s += ',';
+        s += '('; printString(s, i->first);
+        s += ','; printString(s, i->second);
+        s += ')';
+    }
+    
+    s += "])";
+    
+    return s;
 }
 
 

src/libstore/derivations.hh

@@ -1,12 +1,10 @@
 #ifndef __DERIVATIONS_H
 #define __DERIVATIONS_H
 
-#include <aterm1.h>
-
-#include "hash.hh"
-
 #include <map>
 
+#include "types.hh"
+
 
 namespace nix {
 
@@ -53,17 +51,14 @@ struct Derivation
 };
 
 
-/* Hash an aterm. */
-Hash hashTerm(ATerm t);
-
 /* Write a derivation to the Nix store, and return its path. */
 Path writeDerivation(const Derivation & drv, const string & name);
 
 /* Parse a derivation. */
-Derivation parseDerivation(ATerm t);
+Derivation parseDerivation(const string & s);
 
-/* Parse a derivation. */
-ATerm unparseDerivation(const Derivation & drv);
+/* Print a derivation. */
+string unparseDerivation(const Derivation & drv);
 
 /* Check whether a file name ends with the extensions for
    derivations. */

src/libstore/gc.cc

@@ -1,6 +1,5 @@
 #include "globals.hh"
 #include "misc.hh"
-#include "pathlocks.hh"
 #include "local-store.hh"
 
 #include <boost/shared_ptr.hpp>
@@ -31,7 +30,7 @@ static const int defaultGcLevel = 1000;
    read.  To be precise: when they try to create a new temporary root
    file, they will block until the garbage collector has finished /
    yielded the GC lock. */
-static int openGCLock(LockType lockType)
+int LocalStore::openGCLock(LockType lockType)
 {
     Path fnGCLock = (format("%1%/%2%")
         % nixStateDir % gcLockName).str();
@@ -127,7 +126,7 @@ Path addPermRoot(const Path & _storePath, const Path & _gcRoot,
        Instead of reading all the roots, it would be more efficient to
        check if the root is in a directory in or linked from the
        gcroots directory. */
-    if (queryBoolSetting("gc-check-reachability", true)) {
+    if (queryBoolSetting("gc-check-reachability", false)) {
         Roots roots = store->findRoots();
         if (roots.find(gcRoot) == roots.end())
             printMsg(lvlError, 
@@ -136,7 +135,7 @@ Path addPermRoot(const Path & _storePath, const Path & _gcRoot,
                     "therefore, `%2%' might be removed by the garbage collector")
                 % gcRoot % storePath);
     }
-        
+
     /* Grab the global GC root, causing us to block while a GC is in
        progress.  This prevents the set of permanent roots from
        increasing while a GC is in progress. */
@@ -416,18 +415,13 @@ struct LocalStore::GCState
     PathSet busy;
     bool gcKeepOutputs;
     bool gcKeepDerivations;
-
-    bool drvsIndexed;
-    typedef std::multimap<string, Path> DrvsByName;
-    DrvsByName drvsByName; // derivation paths hashed by name attribute
-
-    GCState(GCResults & results_) : results(results_), drvsIndexed(false)
+    GCState(GCResults & results_) : results(results_)
     {
     }
 };
 
 
-static bool doDelete(GCOptions::GCAction action)
+static bool shouldDelete(GCOptions::GCAction action)
 {
     return action == GCOptions::gcDeleteDead
         || action == GCOptions::gcDeleteSpecific;
@@ -441,45 +435,11 @@ bool LocalStore::isActiveTempFile(const GCState & state,
         && state.tempRoots.find(string(path, 0, path.size() - suffix.size())) != state.tempRoots.end();
 }
 
-
-/* Return all the derivations in the Nix store that have `path' as an
-   output.  This function assumes that derivations have the same name
-   as their outputs. */
-PathSet LocalStore::findDerivers(GCState & state, const Path & path)
-{
-    PathSet derivers;
-
-    Path deriver = queryDeriver(path);
-    if (deriver != "") derivers.insert(deriver);
-
-    if (!state.drvsIndexed) {
-        Paths entries = readDirectory(nixStore);
-        foreach (Paths::iterator, i, entries)
-            if (isDerivation(*i))
-                state.drvsByName.insert(std::pair<string, Path>(
-                        getNameOfStorePath(*i), nixStore + "/" + *i));
-        state.drvsIndexed = true;
-    }
-    
-    string name = getNameOfStorePath(path);
-
-    // Urgh, I should have used Haskell...
-    std::pair<GCState::DrvsByName::iterator, GCState::DrvsByName::iterator> range =
-        state.drvsByName.equal_range(name);
-
-    for (GCState::DrvsByName::iterator i = range.first; i != range.second; ++i)
-        if (isValidPath(i->second)) {
-            Derivation drv = derivationFromPath(i->second);
-            foreach (DerivationOutputs::iterator, j, drv.outputs)
-                if (j->second.path == path) derivers.insert(i->second);
-        }
-
-    return derivers;
-}
-
     
 bool LocalStore::tryToDelete(GCState & state, const Path & path)
 {
+    checkInterrupt();
+    
     if (!pathExists(path)) return true;
     if (state.deleted.find(path) != state.deleted.end()) return true;
     if (state.live.find(path) != state.live.end()) return false;
@@ -508,10 +468,10 @@ bool LocalStore::tryToDelete(GCState & state, const Path & path)
            then don't delete the derivation if any of the outputs are
            live. */
         if (state.gcKeepDerivations && isDerivation(path)) {
-            Derivation drv = derivationFromPath(path);
-            foreach (DerivationOutputs::iterator, i, drv.outputs)
-                if (!tryToDelete(state, i->second.path)) {
-                    printMsg(lvlDebug, format("cannot delete derivation `%1%' because its output is alive") % path);
+            PathSet outputs = queryDerivationOutputs(path);
+            foreach (PathSet::iterator, i, outputs)
+                if (!tryToDelete(state, *i)) {
+                    printMsg(lvlDebug, format("cannot delete derivation `%1%' because its output `%2%' is alive") % path % *i);
                     goto isLive;
                 }
         }
@@ -522,18 +482,9 @@ bool LocalStore::tryToDelete(GCState & state, const Path & path)
         if (!pathExists(path)) return true;
 
         /* If gc-keep-outputs is set, then don't delete this path if
-           its deriver is not garbage.  !!! Nix does not reliably
-           store derivers, so we have to look at all derivations to
-           determine which of them derive `path'.  Since this makes
-           the garbage collector very slow to start on large Nix
-           stores, here we just look for all derivations that have the
-           same name as `path' (where the name is the part of the
-           filename after the hash, i.e. the `name' attribute of the
-           derivation).  This is somewhat hacky: currently, the
-           deriver of a path always has the same name as the output,
-           but this might change in the future. */
+           there are derivers of this path that are not garbage. */
         if (state.gcKeepOutputs) {
-            PathSet derivers = findDerivers(state, path);
+            PathSet derivers = queryValidDerivers(path);
             foreach (PathSet::iterator, deriver, derivers) {
                 /* Break an infinite recursion if gc-keep-derivations
                    and gc-keep-outputs are both set by tentatively
@@ -567,7 +518,7 @@ bool LocalStore::tryToDelete(GCState & state, const Path & path)
     }
 
     /* The path is garbage, so delete it. */
-    if (doDelete(state.options.action)) {
+    if (shouldDelete(state.options.action)) {
         printMsg(lvlInfo, format("deleting `%1%'") % path);
 
         unsigned long long bytesFreed, blocksFreed;
@@ -613,6 +564,15 @@ void LocalStore::collectGarbage(const GCOptions & options, GCResults & results)
     
     state.gcKeepOutputs = queryBoolSetting("gc-keep-outputs", false);
     state.gcKeepDerivations = queryBoolSetting("gc-keep-derivations", true);
+
+    /* Using `--ignore-liveness' with `--delete' can have unintended
+       consequences if `gc-keep-outputs' or `gc-keep-derivations' are
+       true (the garbage collector will recurse into deleting the
+       outputs or derivers, respectively).  So disable them. */
+    if (options.action == GCOptions::gcDeleteSpecific && options.ignoreLiveness) {
+        state.gcKeepOutputs = false;
+        state.gcKeepDerivations = false;
+    }
     
     /* Acquire the global GC root.  This prevents
        a) New roots from being added.
@@ -667,7 +627,7 @@ void LocalStore::collectGarbage(const GCOptions & options, GCResults & results)
         vector<Path> entries_(entries.begin(), entries.end());
         random_shuffle(entries_.begin(), entries_.end());
 
-        if (doDelete(state.options.action))
+        if (shouldDelete(state.options.action))
             printMsg(lvlError, format("deleting garbage..."));
         else
             printMsg(lvlError, format("determining live/dead paths..."));

src/libstore/globals.cc

@@ -20,8 +20,9 @@ string nixBinDir = "/UNINIT";
 bool keepFailed = false;
 bool keepGoing = false;
 bool tryFallback = false;
-Verbosity buildVerbosity = lvlInfo;
+Verbosity buildVerbosity = lvlError;
 unsigned int maxBuildJobs = 1;
+unsigned int buildCores = 1;
 bool readOnlyMode = false;
 string thisSystem = "unset";
 time_t maxSilentTime = 0;

src/libstore/globals.hh

@@ -55,6 +55,11 @@ extern Verbosity buildVerbosity;
 /* Maximum number of parallel build jobs.  0 means unlimited. */
 extern unsigned int maxBuildJobs;
 
+/* Number of CPU cores to utilize in parallel within a build, i.e. by passing
+   this number to Make via '-j'. 0 means that the number of actual CPU cores on
+   the local host ought to be auto-detected. */
+extern unsigned int buildCores;
+
 /* Read-only mode.  Don't copy stuff to the store, don't change the
    database. */
 extern bool readOnlyMode;

src/libstore/local-store.hh

@@ -5,6 +5,11 @@
 
 #include "store-api.hh"
 #include "util.hh"
+#include "pathlocks.hh"
+
+
+class sqlite3;
+class sqlite3_stmt;
 
 
 namespace nix {
@@ -12,8 +17,9 @@ namespace nix {
 
 /* Nix store and database schema version.  Version 1 (or 0) was Nix <=
    0.7.  Version 2 was Nix 0.8 and 0.9.  Version 3 is Nix 0.10.
-   Version 4 is Nix 0.11.  Version 5 is Nix 0.12*/
-const int nixSchemaVersion = 5;
+   Version 4 is Nix 0.11.  Version 5 is Nix 0.12-0.16.  Version 6 is
+   Nix 1.0. */
+const int nixSchemaVersion = 6;
 
 
 extern string drvsLogDir;
@@ -41,6 +47,34 @@ struct RunningSubstituter
 };
 
 
+/* Wrapper object to close the SQLite database automatically. */
+struct SQLite
+{
+    sqlite3 * db;
+    SQLite() { db = 0; }
+    ~SQLite();
+    operator sqlite3 * () { return db; }
+};
+
+
+/* Wrapper object to create and destroy SQLite prepared statements. */
+struct SQLiteStmt
+{
+    sqlite3 * db;
+    sqlite3_stmt * stmt;
+    unsigned int curArg;
+    SQLiteStmt() { stmt = 0; }
+    void create(sqlite3 * db, const string & s);
+    void reset();
+    ~SQLiteStmt();
+    operator sqlite3_stmt * () { return stmt; }
+    void bind(const string & value);
+    void bind(int value);
+    void bind64(long long value);
+    void bind();
+};
+    
+
 class LocalStore : public StoreAPI
 {
 private:
@@ -64,6 +98,8 @@ public:
 
     PathSet queryValidPaths();
     
+    ValidPathInfo queryPathInfo(const Path & path);
+
     Hash queryPathHash(const Path & path);
 
     void queryReferences(const Path & path, PathSet & references);
@@ -71,6 +107,14 @@ public:
     void queryReferrers(const Path & path, PathSet & referrers);
 
     Path queryDeriver(const Path & path);
+
+    /* Return all currently valid derivations that have `path' as an
+       output.  (Note that the result of `queryDeriver()' is the
+       derivation that was actually used to produce `path', which may
+       not exist anymore.) */
+    PathSet queryValidDerivers(const Path & path);
+
+    PathSet queryDerivationOutputs(const Path & path);
     
     PathSet querySubstitutablePaths();
     
@@ -132,8 +176,7 @@ public:
        execution of the derivation (or something equivalent).  Also
        register the hash of the file system contents of the path.  The
        hash must be a SHA-256 hash. */
-    void registerValidPath(const Path & path,
-        const Hash & hash, const PathSet & references, const Path & deriver);
+    void registerValidPath(const ValidPathInfo & info);
 
     void registerValidPaths(const ValidPathInfos & infos);
 
@@ -144,6 +187,10 @@ public:
     /* Query whether `path' previously failed to build. */
     bool hasPathFailed(const Path & path);
 
+    PathSet queryFailedPaths();
+
+    void clearFailedPaths(const PathSet & paths);
+
 private:
 
     Path schemaPath;
@@ -151,45 +198,63 @@ private:
     /* Lock file used for upgrading. */
     AutoCloseFD globalLock;
 
-    /* !!! The cache can grow very big.  Maybe it should be pruned
-       every once in a while. */
-    std::map<Path, ValidPathInfo> pathInfoCache;
+    /* The SQLite database object. */
+    SQLite db;
 
-    /* Store paths for which the referrers file must be purged. */
-    PathSet delayedUpdates;
-
-    /* Whether to do an fsync() after writing Nix metadata. */
-    bool doFsync;
+    /* Some precompiled SQLite statements. */
+    SQLiteStmt stmtRegisterValidPath;
+    SQLiteStmt stmtUpdatePathInfo;
+    SQLiteStmt stmtAddReference;
+    SQLiteStmt stmtQueryPathInfo;
+    SQLiteStmt stmtQueryReferences;
+    SQLiteStmt stmtQueryReferrers;
+    SQLiteStmt stmtInvalidatePath;
+    SQLiteStmt stmtRegisterFailedPath;
+    SQLiteStmt stmtHasPathFailed;
+    SQLiteStmt stmtQueryFailedPaths;
+    SQLiteStmt stmtClearFailedPath;
+    SQLiteStmt stmtAddDerivationOutput;
+    SQLiteStmt stmtQueryValidDerivers;
+    SQLiteStmt stmtQueryDerivationOutputs;
 
     int getSchema();
 
-    void registerValidPath(const ValidPathInfo & info, bool ignoreValidity = false);
+    void openDB(bool create);
 
-    ValidPathInfo queryPathInfo(const Path & path, bool ignoreErrors = false);
+    unsigned long long queryValidPathId(const Path & path);
 
+    unsigned long long addValidPath(const ValidPathInfo & info);
+        
+    void addReference(unsigned long long referrer, unsigned long long reference);
+    
     void appendReferrer(const Path & from, const Path & to, bool lock);
     
     void rewriteReferrers(const Path & path, bool purge, PathSet referrers);
 
-    void flushDelayedUpdates();
-    
-    bool queryReferrersInternal(const Path & path, PathSet & referrers);
-    
     void invalidatePath(const Path & path);
-    
-    void upgradeStore12();
+
+    void verifyPath(const Path & path, const PathSet & store,
+        PathSet & done, PathSet & validPaths);
+
+    void updatePathInfo(const ValidPathInfo & info);
+
+    void upgradeStore6();
+    PathSet queryValidPathsOld();
+    ValidPathInfo queryPathInfoOld(const Path & path);
 
     struct GCState;
 
     bool tryToDelete(GCState & state, const Path & path);
     
-    PathSet findDerivers(GCState & state, const Path & path);
-    
     bool isActiveTempFile(const GCState & state,
         const Path & path, const string & suffix);
         
+    int openGCLock(LockType lockType);
+    
     void startSubstituter(const Path & substituter,
         RunningSubstituter & runningSubstituter);
+
+    Path createTempDirInStore();
 };
 
 

src/libstore/misc.cc

@@ -2,8 +2,6 @@
 #include "store-api.hh"
 #include "local-store.hh"
 
-#include <aterm2.h>
-
 
 namespace nix {
 
@@ -12,9 +10,7 @@ Derivation derivationFromPath(const Path & drvPath)
 {
     assertStorePath(drvPath);
     store->ensurePath(drvPath);
-    ATerm t = ATreadFromNamedFile(drvPath.c_str());
-    if (!t) throw Error(format("cannot read aterm from `%1%'") % drvPath);
-    return parseDerivation(t);
+    return parseDerivation(readFile(drvPath));
 }
 
 
@@ -31,10 +27,10 @@ void computeFSClosure(const Path & storePath,
         store->queryReferences(storePath, references);
 
     if (includeOutputs && isDerivation(storePath)) {
-        Derivation drv = derivationFromPath(storePath);
-        foreach (DerivationOutputs::iterator, i, drv.outputs)
-            if (store->isValidPath(i->second.path))
-                computeFSClosure(i->second.path, paths, flipDirection, true);
+        PathSet outputs = store->queryDerivationOutputs(storePath);
+        foreach (PathSet::iterator, i, outputs)
+            if (store->isValidPath(*i))
+                computeFSClosure(*i, paths, flipDirection, true);
     }
 
     foreach (PathSet::iterator, i, references)
@@ -52,9 +48,9 @@ Path findOutput(const Derivation & drv, string id)
 
 void queryMissing(const PathSet & targets,
     PathSet & willBuild, PathSet & willSubstitute, PathSet & unknown,
-    unsigned long long & downloadSize)
+    unsigned long long & downloadSize, unsigned long long & narSize)
 {
-    downloadSize = 0;
+    downloadSize = narSize = 0;
     
     PathSet todo(targets.begin(), targets.end()), done;
 
@@ -92,6 +88,7 @@ void queryMissing(const PathSet & targets,
             if (store->querySubstitutablePathInfo(p, info)) {
                 willSubstitute.insert(p);
                 downloadSize += info.downloadSize;
+                narSize += info.narSize;
                 todo.insert(info.references.begin(), info.references.end());
             } else
                 unknown.insert(p);

src/libstore/misc.hh

@@ -31,7 +31,7 @@ Path findOutput(const Derivation & drv, string id);
    will be substituted. */
 void queryMissing(const PathSet & targets,
     PathSet & willBuild, PathSet & willSubstitute, PathSet & unknown,
-    unsigned long long & downloadSize);
+    unsigned long long & downloadSize, unsigned long long & narSize);
 
 
 }

src/libstore/optimise-store.cc

@@ -68,7 +68,7 @@ static void hashAndLink(bool dryRun, HashToPath & hashToPath,
            the contents of the symlink (i.e. the result of
            readlink()), not the contents of the target (which may not
            even exist). */
-        Hash hash = hashPath(htSHA256, path);
+        Hash hash = hashPath(htSHA256, path).first;
         stats.totalFiles++;
         printMsg(lvlDebug, format("`%1%' has hash `%2%'") % path % printHash(hash));
 
@@ -119,9 +119,23 @@ static void hashAndLink(bool dryRun, HashToPath & hashToPath,
             }
 
             /* Atomically replace the old file with the new hard link. */
-            if (rename(tempLink.c_str(), path.c_str()) == -1)
+            if (rename(tempLink.c_str(), path.c_str()) == -1) {
+                if (errno == EMLINK) {
+                    /* Some filesystems generate too many links on the
+                       rename, rather than on the original link.
+                       (Probably it temporarily increases the st_nlink
+                       field before decreasing it again.) */
+                    printMsg(lvlInfo, format("`%1%' has maximum number of links") % prevPath.first);
+                    hashToPath[hash] = std::pair<Path, ino_t>(path, st.st_ino);
+
+                    /* Unlink the temp link. */
+                    if (unlink(tempLink.c_str()) == -1)
+                        printMsg(lvlError, format("unable to unlink `%1%'") % tempLink);
+                    return;
+                }
                 throw SysError(format("cannot rename `%1%' to `%2%'")
                     % tempLink % path);
+            }
         } else
             printMsg(lvlTalkative, format("would link `%1%' to `%2%'") % path % prevPath.first);
         

src/libstore/references.cc

@@ -81,7 +81,7 @@ void RefScanSink::operator () (const unsigned char * data, unsigned int len)
 
 
 PathSet scanForReferences(const string & path,
-    const PathSet & refs, Hash & hash)
+    const PathSet & refs, HashResult & hash)
 {
     RefScanSink sink;
     std::map<string, Path> backMap;

src/libstore/references.hh

@@ -7,7 +7,7 @@
 namespace nix {
 
 PathSet scanForReferences(const Path & path, const PathSet & refs,
-    Hash & hash);
+    HashResult & hash);
     
 }
 

src/libstore/remote-store.cc

@@ -13,6 +13,7 @@
 
 #include <iostream>
 #include <unistd.h>
+#include <cstring>
 
 
 namespace nix {
@@ -96,10 +97,6 @@ void RemoteStore::forkSlave()
     if (worker == "")
         worker = nixBinDir + "/nix-worker";
 
-    string verbosityArg = "-";
-    for (int i = 1; i < verbosity; ++i)
-        verbosityArg += "v";
-
     child = fork();
     
     switch (child) {
@@ -119,10 +116,7 @@ void RemoteStore::forkSlave()
             close(fdSocket);
             close(fdChild);
 
-            execlp(worker.c_str(), worker.c_str(), "--slave",
-                /* hacky - must be at the end */
-                verbosityArg == "-" ? NULL : verbosityArg.c_str(),
-                NULL);
+            execlp(worker.c_str(), worker.c_str(), "--slave", NULL);
 
             throw SysError(format("executing `%1%'") % worker);
             
@@ -158,6 +152,7 @@ void RemoteStore::connectToDaemon()
     addr.sun_family = AF_UNIX;
     if (socketPathRel.size() >= sizeof(addr.sun_path))
         throw Error(format("socket path `%1%' is too long") % socketPathRel);
+    using namespace std;
     strcpy(addr.sun_path, socketPathRel.c_str());
     
     if (connect(fdSocket, (struct sockaddr *) &addr, sizeof(addr)) == -1)
@@ -196,6 +191,8 @@ void RemoteStore::setOptions()
         writeInt(logType, to);
         writeInt(printBuildTrace, to);
     }
+    if (GET_PROTOCOL_MINOR(daemonVersion) >= 6)
+        writeInt(buildCores, to);
     processStderr();
 }
 
@@ -214,7 +211,9 @@ bool RemoteStore::isValidPath(const Path & path)
 PathSet RemoteStore::queryValidPaths()
 {
     openConnection();
-    throw Error("not implemented");
+    writeInt(wopQueryValidPaths, to);
+    processStderr();
+    return readStorePaths(from);
 }
 
 
@@ -243,10 +242,29 @@ bool RemoteStore::querySubstitutablePathInfo(const Path & path,
     if (info.deriver != "") assertStorePath(info.deriver);
     info.references = readStorePaths(from);
     info.downloadSize = readLongLong(from);
+    info.narSize = GET_PROTOCOL_MINOR(daemonVersion) >= 7 ? readLongLong(from) : 0;
     return true;
 }
 
 
+ValidPathInfo RemoteStore::queryPathInfo(const Path & path)
+{
+    openConnection();
+    writeInt(wopQueryPathInfo, to);
+    writeString(path, to);
+    processStderr();
+    ValidPathInfo info;
+    info.path = path;
+    info.deriver = readString(from);
+    if (info.deriver != "") assertStorePath(info.deriver);
+    info.hash = parseHash(htSHA256, readString(from));
+    info.references = readStorePaths(from);
+    info.registrationTime = readInt(from);
+    info.narSize = readLongLong(from);
+    return info;
+}
+
+
 Hash RemoteStore::queryPathHash(const Path & path)
 {
     openConnection();
@@ -294,6 +312,16 @@ Path RemoteStore::queryDeriver(const Path & path)
 }
 
 
+PathSet RemoteStore::queryDerivationOutputs(const Path & path)
+{
+    openConnection();
+    writeInt(wopQueryDerivationOutputs, to);
+    writeString(path, to);
+    processStderr();
+    return readStorePaths(from);
+}
+
+
 Path RemoteStore::addToStore(const Path & _srcPath,
     bool recursive, HashType hashAlgo, PathFilter & filter)
 {
@@ -439,6 +467,25 @@ void RemoteStore::collectGarbage(const GCOptions & options, GCResults & results)
 }
 
 
+PathSet RemoteStore::queryFailedPaths()
+{
+    openConnection();
+    writeInt(wopQueryFailedPaths, to);
+    processStderr();
+    return readStorePaths(from);
+}
+
+
+void RemoteStore::clearFailedPaths(const PathSet & paths)
+{
+    openConnection();
+    writeInt(wopClearFailedPaths, to);
+    writeStringSet(paths, to);
+    processStderr();
+    readInt(from);
+}
+
+
 void RemoteStore::processStderr(Sink * sink, Source * source)
 {
     unsigned int msg;
@@ -462,8 +509,11 @@ void RemoteStore::processStderr(Sink * sink, Source * source)
             writeToStderr((const unsigned char *) s.c_str(), s.size());
         }
     }
-    if (msg == STDERR_ERROR)
-        throw Error(readString(from));
+    if (msg == STDERR_ERROR) {
+        string error = readString(from);
+        unsigned int status = GET_PROTOCOL_MINOR(daemonVersion) >= 8 ? readInt(from) : 1;
+        throw Error(error, status);
+    }
     else if (msg != STDERR_LAST)
         throw Error("protocol error processing standard error");
 }

src/libstore/remote-store.hh

@@ -29,6 +29,8 @@ public:
 
     PathSet queryValidPaths();
     
+    ValidPathInfo queryPathInfo(const Path & path);
+
     Hash queryPathHash(const Path & path);
 
     void queryReferences(const Path & path, PathSet & references);
@@ -37,6 +39,8 @@ public:
 
     Path queryDeriver(const Path & path);
     
+    PathSet queryDerivationOutputs(const Path & path);
+    
     bool hasSubstitutes(const Path & path);
     
     bool querySubstitutablePathInfo(const Path & path,
@@ -68,6 +72,10 @@ public:
 
     void collectGarbage(const GCOptions & options, GCResults & results);
     
+    PathSet queryFailedPaths();
+
+    void clearFailedPaths(const PathSet & paths);
+    
 private:
     AutoCloseFD fdSocket;
     FdSink to;

src/libstore/schema.sql

@@ -0,0 +1,44 @@
+create table if not exists ValidPaths (
+    id               integer primary key autoincrement not null,
+    path             text unique not null,
+    hash             text not null,
+    registrationTime integer not null,
+    deriver          text,
+    narSize          integer
+);
+
+create table if not exists Refs (
+    referrer  integer not null,
+    reference integer not null,
+    primary key (referrer, reference),
+    foreign key (referrer) references ValidPaths(id) on delete cascade,
+    foreign key (reference) references ValidPaths(id) on delete restrict
+);
+
+create index if not exists IndexReferrer on Refs(referrer);
+create index if not exists IndexReference on Refs(reference);
+
+-- Paths can refer to themselves, causing a tuple (N, N) in the Refs
+-- table.  This causes a deletion of the corresponding row in
+-- ValidPaths to cause a foreign key constraint violation (due to `on
+-- delete restrict' on the `reference' column).  Therefore, explicitly
+-- get rid of self-references.
+create trigger if not exists DeleteSelfRefs before delete on ValidPaths
+  begin
+    delete from Refs where referrer = old.id and reference = old.id;
+  end;
+
+create table if not exists DerivationOutputs (
+    drv  integer not null,
+    id   text not null, -- symbolic output id, usually "out"
+    path text not null,
+    primary key (drv, id),
+    foreign key (drv) references ValidPaths(id) on delete cascade
+);
+
+create index if not exists IndexDerivationOutputs on DerivationOutputs(path);
+
+create table if not exists FailedPaths (
+    path text primary key not null,
+    time integer not null
+);

src/libstore/store-api.cc

@@ -1,7 +1,6 @@
 #include "store-api.hh"
 #include "globals.hh"
 #include "util.hh"
-#include "derivations.hh"
 
 #include <limits.h>
 
@@ -53,18 +52,6 @@ Path toStorePath(const Path & path)
 }
 
 
-string getNameOfStorePath(const Path & path)
-{
-    Path::size_type slash = path.rfind('/');
-    string p = slash == Path::npos ? path : string(path, slash + 1);
-    Path::size_type dash = p.find('-');
-    assert(dash != Path::npos);
-    string p2 = string(p, dash + 1);
-    if (isDerivation(p2)) p2 = string(p2, 0, p2.size() - 4);
-    return p2;
-}
-
-
 Path followLinksToStore(const Path & _path)
 {
     Path path = absPath(_path);
@@ -203,7 +190,7 @@ std::pair<Path, Hash> computeStorePathForPath(const Path & srcPath,
     bool recursive, HashType hashAlgo, PathFilter & filter)
 {
     HashType ht(hashAlgo);
-    Hash h = recursive ? hashPath(ht, srcPath, filter) : hashFile(ht, srcPath);
+    Hash h = recursive ? hashPath(ht, srcPath, filter).first : hashFile(ht, srcPath);
     string name = baseNameOf(srcPath);
     Path dstPath = makeFixedOutputPath(recursive, hashAlgo, h, name);
     return std::pair<Path, Hash>(dstPath, h);
@@ -229,7 +216,7 @@ Path computeStorePathForText(const string & name, const string & s,
 /* Return a string accepted by decodeValidPathInfo() that
    registers the specified paths as valid.  Note: it's the
    responsibility of the caller to provide a closure. */
-string makeValidityRegistration(const PathSet & paths,
+string StoreAPI::makeValidityRegistration(const PathSet & paths,
     bool showDerivers, bool showHash)
 {
     string s = "";
@@ -237,18 +224,19 @@ string makeValidityRegistration(const PathSet & paths,
     foreach (PathSet::iterator, i, paths) {
         s += *i + "\n";
 
-        if (showHash)
-            s += printHash(store->queryPathHash(*i)) + "\n";
+        ValidPathInfo info = queryPathInfo(*i);
 
-        Path deriver = showDerivers ? store->queryDeriver(*i) : "";
+        if (showHash) {
+            s += printHash(info.hash) + "\n";
+            s += (format("%1%\n") % info.narSize).str();
+        }
+
+        Path deriver = showDerivers ? info.deriver : "";
         s += deriver + "\n";
 
-        PathSet references;
-        store->queryReferences(*i, references);
+        s += (format("%1%\n") % info.references.size()).str();
 
-        s += (format("%1%\n") % references.size()).str();
-            
-        foreach (PathSet::iterator, j, references)
+        foreach (PathSet::iterator, j, info.references)
             s += *j + "\n";
     }
 
@@ -265,6 +253,8 @@ ValidPathInfo decodeValidPathInfo(std::istream & str, bool hashGiven)
         string s;
         getline(str, s);
         info.hash = parseHash(htSHA256, s);
+        getline(str, s);
+        if (!string2Int(s, info.narSize)) throw Error("number expected");
     }
     getline(str, info.deriver);
     string s; int n;

src/libstore/store-api.hh

@@ -1,14 +1,14 @@
 #ifndef __STOREAPI_H
 #define __STOREAPI_H
 
+#include "hash.hh"
+#include "serialise.hh"
+
 #include <string>
 #include <map>
 
 #include <boost/shared_ptr.hpp>
 
-#include "hash.hh"
-#include "serialise.hh"
-
 
 namespace nix {
 
@@ -16,8 +16,6 @@ namespace nix {
 typedef std::map<Path, Path> Roots;
 
 
-
-
 struct GCOptions
 {
     /* Garbage collector operation:
@@ -89,9 +87,25 @@ struct SubstitutablePathInfo
     Path deriver;
     PathSet references;
     unsigned long long downloadSize; /* 0 = unknown or inapplicable */
+    unsigned long long narSize; /* 0 = unknown */
 };
 
 
+struct ValidPathInfo 
+{
+    Path path;
+    Path deriver;
+    Hash hash;
+    PathSet references;
+    time_t registrationTime;
+    unsigned long long narSize; // 0 = unknown
+    unsigned long long id; // internal use only
+    ValidPathInfo() : registrationTime(0), narSize(0) { }
+};
+
+typedef list<ValidPathInfo> ValidPathInfos;
+
+
 class StoreAPI 
 {
 public:
@@ -104,6 +118,9 @@ public:
     /* Query the set of valid paths. */
     virtual PathSet queryValidPaths() = 0;
 
+    /* Query information about a valid path. */
+    virtual ValidPathInfo queryPathInfo(const Path & path) = 0;
+
     /* Queries the hash of a valid path. */ 
     virtual Hash queryPathHash(const Path & path) = 0;
 
@@ -112,33 +129,18 @@ public:
     virtual void queryReferences(const Path & path,
         PathSet & references) = 0;
 
-    /* Like queryReferences, but with self-references filtered out. */
-    PathSet queryReferencesNoSelf(const Path & path)
-    {
-        PathSet res;
-        queryReferences(path, res);
-        res.erase(path);
-        return res;
-    }
-
     /* Queries the set of incoming FS references for a store path.
        The result is not cleared. */
     virtual void queryReferrers(const Path & path,
         PathSet & referrers) = 0;
 
-    /* Like queryReferrers, but with self-references filtered out. */
-    PathSet queryReferrersNoSelf(const Path & path)
-    {
-        PathSet res;
-        queryReferrers(path, res);
-        res.erase(path);
-        return res;
-    }
-
     /* Query the deriver of a store path.  Return the empty string if
        no deriver has been set. */
     virtual Path queryDeriver(const Path & path) = 0;
 
+    /* Query the outputs of the derivation denoted by `path'. */
+    virtual PathSet queryDerivationOutputs(const Path & path) = 0;
+    
     /* Query whether a path has substitutes. */
     virtual bool hasSubstitutes(const Path & path) = 0;
 
@@ -224,6 +226,19 @@ public:
 
     /* Perform a garbage collection. */
     virtual void collectGarbage(const GCOptions & options, GCResults & results) = 0;
+
+    /* Return the set of paths that have failed to build.*/
+    virtual PathSet queryFailedPaths() = 0;
+
+    /* Clear the "failed" status of the given paths.  The special
+       value `*' causes all failed paths to be cleared. */
+    virtual void clearFailedPaths(const PathSet & paths) = 0;
+
+    /* Return a string representing information about the path that
+       can be loaded into the database using `nix-store --load-db' or
+       `nix-store --register-validity'. */
+    string makeValidityRegistration(const PathSet & paths,
+        bool showDerivers, bool showHash);
 };
 
 
@@ -243,12 +258,6 @@ void checkStoreName(const string & name);
 Path toStorePath(const Path & path);
 
 
-/* Get the "name" part of a store path, that is, the part after the
-   hash and the dash, and with any ".drv" suffix removed
-   (e.g. /nix/store/<hash>-foo-1.2.3.drv => foo-1.2.3). */
-string getNameOfStorePath(const Path & path);
-
-
 /* Follow symlinks until we end up with a path in the Nix store. */
 Path followLinksToStore(const Path & path);
 
@@ -323,21 +332,6 @@ boost::shared_ptr<StoreAPI> openStore();
 string showPaths(const PathSet & paths);
 
 
-string makeValidityRegistration(const PathSet & paths,
-    bool showDerivers, bool showHash);
-    
-struct ValidPathInfo 
-{
-    Path path;
-    Path deriver;
-    Hash hash;
-    PathSet references;
-    time_t registrationTime;
-    ValidPathInfo() : registrationTime(0) { }
-};
-
-typedef list<ValidPathInfo> ValidPathInfos;
-
 ValidPathInfo decodeValidPathInfo(std::istream & str,
     bool hashGiven = false);
 

src/libstore/worker-protocol.hh

@@ -8,7 +8,7 @@ namespace nix {
 #define WORKER_MAGIC_1 0x6e697863
 #define WORKER_MAGIC_2 0x6478696f
 
-#define PROTOCOL_VERSION 0x105
+#define PROTOCOL_VERSION 0x108
 #define GET_PROTOCOL_MAJOR(x) ((x) & 0xff00)
 #define GET_PROTOCOL_MINOR(x) ((x) & 0x00ff)
 
@@ -34,6 +34,11 @@ typedef enum {
     wopSetOptions = 19,
     wopCollectGarbage = 20,
     wopQuerySubstitutablePathInfo = 21,
+    wopQueryDerivationOutputs = 22,
+    wopQueryValidPaths = 23,
+    wopQueryFailedPaths = 24,
+    wopClearFailedPaths = 25,
+    wopQueryPathInfo = 26,
 } WorkerOp;
 
 

src/libterm/grammar.hh

@@ -1,112 +0,0 @@
-// This file contains macro used to define the grammar.  You should use
-// these to define the macro named TRM_GRAMMAR_NODES(Interface, Final)
-// before inluding any of the libterm files.
-
-#ifndef _LIBTERM_GRAMMAR_HH
-# define _LIBTERM_GRAMMAR_HH
-
-# include "pp.hh"
-
-
-// These macro are used to define how to manipulate each argument.  If the
-// argument should be given by reference or by copy.  You should use
-// references when the element goes over a specific size and if you accept
-// to see it living on the stack.  You should use copies when the element is
-// small and if you prefer to have it inside a register.
-
-# define TRM_TYPE_REF(Type, Name) (const Type&, Type &, Name)
-# define TRM_TYPE_COPY(Type, Name) (const Type, Type &, Name)
-# define TRM_TYPE_TERM(Type, Name) TRM_TYPE_COPY(A ## Type, Name)
-
-
-// These macro are used as shortcuts for the declaration of common type of
-// grammar nodes.
-
-# define TRM_GRAMMAR_NODE_BINOP(Final, Name)                    \
-  Final(                                                        \
-    Name, Expr,                                                 \
-    (2, (TRM_TYPE_TERM(Expr, lhs), TRM_TYPE_TERM(Expr, rhs))),  \
-    (0,())                                                      \
-  )
-
-# define TRM_GRAMMAR_NODE_SINGLETON(Final, Name)        \
-  Final(Name, Expr, (0,()), (0,()))
-
-
-// Handle the different usage of the variables and attributes.  Arguments
-// are suffixed with a '_' and attributes are not.  These macro are
-// expecting the result of the TRM_TYPE macros as argument.
-
-# define TRM_CONST_ABSTRACT_COPY_DECL_(Copy, Ref, Name) Copy Name;
-# define TRM_CONST_ABSTRACT_COPY_ARG_(Copy, Ref, Name)  Copy Name ## _
-# define TRM_ABSTRACT_REF_ARG_(Copy, Ref, Name)  Ref Name ## _
-# define TRM_INIT_ATTRIBUTES_(Copy, Ref, Name)  Name (Name ## _)
-# define TRM_ARGUMENTS_(Copy, Ref, Name)  Name ## _
-# define TRM_COPY_PTR_ATTR_IN_ARG_(Copy, Ref, Name)  Name ## _ = ptr-> Name;
-# define TRM_LESS_RHS_OR_(Copy, Ref, Name)  Name < arg_rhs. Name ||
-
-// These macro are shortcuts used to remove extra parenthesies added by
-// TRM_TYPE_* macros.  Without such parenthesies TRM_APPLY_HELPER won't be
-// able to give the argument to these macros and arrays won't be well
-// formed.
-
-# define TRM_CONST_ABSTRACT_COPY_DECL(Elt) TRM_CONST_ABSTRACT_COPY_DECL_ Elt
-# define TRM_CONST_ABSTRACT_COPY_ARG(Elt) TRM_CONST_ABSTRACT_COPY_ARG_ Elt
-# define TRM_ABSTRACT_REF_ARG(Elt) TRM_ABSTRACT_REF_ARG_ Elt
-# define TRM_INIT_ATTRIBUTES(Elt) TRM_INIT_ATTRIBUTES_ Elt
-# define TRM_ARGUMENTS(Elt) TRM_ARGUMENTS_ Elt
-# define TRM_COPY_PTR_ATTR_IN_ARG(Elt) TRM_COPY_PTR_ATTR_IN_ARG_ Elt
-# define TRM_LESS_RHS_OR(Elt) TRM_LESS_RHS_OR_ Elt
-
-
-
-/*
-// Test case.  This should be moved inside the libexpr or generated in
-// another manner.
-# define TRM_GAMMAR_NODES(Interface, Final)
-  Interface(Loc, Term, TRM_NIL, TRM_NIL)
-  Interface(Pattern, Term, TRM_NIL, TRM_NIL)
-  Interface(Expr, Term, TRM_NIL, TRM_NIL)
-  Final(Pos, Loc, (TRM_TYPE_REF(std::string, file)
-                   TRM_TYPE_COPY(int, line)
-                   TRM_TYPE_COPY(int, column)), TRM_NIL)
-  Final(NoPos, Loc, TRM_NIL, TRM_NIL)
-  Final(String, Expr, (TRM_TYPE_REF(std::string, value)), TRM_NIL)
-  Final(Var, Expr, (TRM_TYPE_REF(std::string, name)), TRM_NIL)
-  Final(Path, Expr, (TRM_TYPE_REF(std::string, filename)), TRM_NIL)
-  Final(Int, Expr, (TRM_TYPE_COPY(int, value)), TRM_NIL)
-  Final(Function, Expr, (TRM_TYPE_TERM(Pattern, pattern)
-                         TRM_TYPE_TERM(Expr, body)
-                         TRM_TYPE_TERM(Pos, position)), TRM_NIL)
-  Final(Assert, Expr, (TRM_TYPE_TERM(Expr, cond)
-                       TRM_TYPE_TERM(Expr, body)
-                       TRM_TYPE_TERM(Pos, position)), TRM_NIL)
-  Final(With, Expr, (TRM_TYPE_TERM(Expr, set)
-                     TRM_TYPE_TERM(Expr, body)
-                     TRM_TYPE_TERM(Pos, position)), TRM_NIL)
-  Final(If, Expr, (TRM_TYPE_TERM(Expr, cond)
-                   TRM_TYPE_TERM(Expr, thenPart)
-                   TRM_TYPE_TERM(Expr, elsePart)), TRM_NIL)
-  Final(OpNot, Expr, (TRM_TYPE_TERM(Expr, cond)), TRM_NIL)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpEq)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpNEq)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpAnd)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpOr)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpImpl)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpUpdate)
-  TRM_GRAMMAR_NODE_BINOP(Final, SubPath)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpHasAttr)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpPlus)
-  TRM_GRAMMAR_NODE_BINOP(Final, OpConcat)
-  Final(Call, Expr, (TRM_TYPE_TERM(Expr, function)
-                     TRM_TYPE_TERM(Expr, argument)), TRM_NIL)
-  Final(Select, Expr, (TRM_TYPE_TERM(Expr, set)
-                       TRM_TYPE_TERM(Var, var)), TRM_NIL)
-  Final(BlackHole, Expr, TRM_NIL, TRM_NIL)
-  Final(Undefined, Expr, TRM_NIL, TRM_NIL)
-  Final(Removed, Expr, TRM_NIL, TRM_NIL)
-
-*/
-
-
-#endif

src/libterm/pp.hh

@@ -1,94 +0,0 @@
-
-#ifndef _LIBTERM_PP_HH
-# define _LIBTERM_PP_HH
-
-# include <boost/preprocessor/tuple.hpp>
-# include <boost/preprocessor/seq.hpp>
-# include <boost/preprocessor/cat.hpp>
-# include <boost/preprocessor/array.hpp>
-
-// a few shortcuts to keep things clear.
-
-# define TRM_NIL
-# define TRM_NIL_ARGS(...)
-# define TRM_EMPTY_ARRAY  (0, ())
-
-# define TRM_SEQ_HEAD BOOST_PP_SEQ_HEAD
-# define TRM_SEQ_TAIL BOOST_PP_SEQ_TAIL
-
-# define TRM_ARRAY_DATA BOOST_PP_ARRAY_DATA
-# define TRM_ARRAY_SIZE BOOST_PP_ARRAY_SIZE
-# define TRM_ARRAY_ELEM BOOST_PP_ARRAY_ELEM
-# define TRM_SEQ_TO_ARRAY BOOST_PP_SEQ_TO_ARRAY
-# define TRM_ARRAY_TO_SEQ(Array)                                        \
-  BOOST_PP_TUPLE_TO_SEQ(TRM_ARRAY_SIZE(Array), TRM_ARRAY_DATA(Array))
-
-// TRM_EVAL macros are used to do the convertion of array to sequence and to
-// provide a default case when the array cannot be converted into a
-// sequence.  This case happens when you need to convert a zero size array.
-
-# define TRM_EVAL_0(Macro1, Macro2, Array, Default) Default
-# define TRM_EVAL_1(Macro1, Macro2, Array, Default)     \
-  Macro1(Macro2, TRM_ARRAY_TO_SEQ(Array))
-
-# define TRM_EVAL_2 TRM_EVAL_1
-# define TRM_EVAL_3 TRM_EVAL_1
-# define TRM_EVAL_4 TRM_EVAL_1
-
-# define TRM_EVAL_II(eval) eval
-# define TRM_EVAL_I(n, _) TRM_EVAL_ ## n
-# define TRM_EVAL(Macro1, Macro2, Array, Default)                       \
-  TRM_EVAL_II(TRM_EVAL_I Array (Macro1, Macro2, Array, Default))
-
-# define TRM_DEFAULT1(D0)  D0
-
-// Default TRM_<fun> are taking an array and return a sequence.  To keep the
-// same type, you need to specify which type you expect to have.  The reason
-// is that array support to have no values where sequence do not support it.
-// On the contrary, all operations are well defined on sequences but none
-// are defined on arrays.
-
-# define TRM_MAP_HELPER(R, Macro, Elt) (Macro(Elt))
-# define TRM_MAP_(Macro, Seq)                           \
-  BOOST_PP_SEQ_FOR_EACH(TRM_MAP_HELPER, Macro, Seq)
-# define TRM_MAP_SEQ(Macro, Seq)                \
-  TRM_MAP_(Macro, Seq)
-# define TRM_MAP_ARRAY_(Macro, Seq)             \
-  TRM_SEQ_TO_ARRAY(TRM_MAP_SEQ(Macro, Seq))
-# define TRM_MAP_ARRAY(Macro, Array)            \
-  TRM_EVAL(TRM_MAP_ARRAY_, Macro, Array,        \
-    TRM_DEFAULT1(TRM_EMPTY_ARRAY)               \
-  )
-# define TRM_MAP(Macro, Array)                  \
-  TRM_EVAL(TRM_MAP_SEQ, Macro, Array,           \
-    TRM_DEFAULT1(TRM_NIL)                       \
-  )
-
-// Apply a macro on all elements (array / sequence)
-
-# define TRM_APPLY_HELPER(R, Macro, Elt) Macro(Elt)
-# define TRM_APPLY_(Macro, Seq)                         \
-  BOOST_PP_SEQ_FOR_EACH(TRM_APPLY_HELPER, Macro, Seq)
-# define TRM_APPLY_SEQ(Macro, Seq)              \
-  TRM_APPLY_(Macro, Seq)
-# define TRM_APPLY(Macro, Array)                \
-  TRM_EVAL(TRM_APPLY_, Macro, Array,            \
-    TRM_DEFAULT1(TRM_NIL)                       \
-  )
-
-// Apply a macro on all elements (array / sequence) and separate them by a
-// comma.
-
-# define TRM_SEPARATE_COMMA_HELPER(Elt) , Elt
-# define TRM_SEPARATE_COMMA_(_, Seq)                            \
-  TRM_SEQ_HEAD(Seq)                                             \
-  TRM_APPLY_SEQ(TRM_SEPARATE_COMMA_HELPER, TRM_SEQ_TAIL(Seq))
-# define TRM_SEPARATE_COMMA_SEQ(Seq)            \
-  TRM_SEPARATE_COMMA(TRM_SEQ_TO_ARRAY(Seq))
-# define TRM_SEPARATE_COMMA(Array)                      \
-  TRM_EVAL(TRM_SEPARATE_COMMA_, dummy, Array,           \
-    TRM_DEFAULT1(TRM_NIL)     \
-  )
-
-
-#endif

src/libterm/term.hh

@@ -1,202 +0,0 @@
-
-#ifndef _LIBTERM_TERM_CLASS_FWD_HH
-# define _LIBTERM_TERM_CLASS_FWD_HH
-
-# include <set>
-# include "visitor_fwd.hh"
-
-namespace term
-{
-  // Performance issue: Abstract classes should be copied where fully
-  // defined terms should be used with constant references.  An ATerm is
-  // only composed by a pointer and should not use any virtual method.  By
-  // following this rule, the compiler can optimize the size to the size of
-  // its members (which is a pointer).
-
-  class ATermImpl
-  {
-  public:
-    ATermImpl();
-    virtual ~ATermImpl();
-
-  public:
-    virtual ATerm accept(ATermVisitor& v) const = 0;
-  };
-
-
-  class ATerm
-  {
-  public:
-    ATerm(const ATerm& t);
-    ATerm();
-
-  protected:
-    ATerm(const ATermImpl* ptr);
-
-  public:
-    ATerm accept(ATermVisitor& v) const;
-
-  public:
-    bool operator== (const ATerm& rhs) const;
-    bool operator!= (const ATerm& rhs) const;
-    bool operator <(const ATerm& rhs) const;
-    operator bool();
-
-  public:
-    const ATermImpl* get_ptr() const;
-
-  protected:
-    const ATermImpl* ptr_;
-  };
-
-
-  struct Term : public ATermImpl
-  {
-  public:
-    typedef Term this_type;
-    typedef ATerm term;
-
-  public:
-    bool operator <(const this_type&) const;
-  };
-
-
-  // [
-
-// Declare the constructor
-# define TRM_IMPL_CTR_DECL(Name, Base, Attributes, BaseArgs)            \
-    Name ( TRM_SEPARATE_COMMA(                                          \
-             TRM_MAP_ARRAY(TRM_CONST_ABSTRACT_COPY_ARG, Attributes)     \
-         ) );
-
-// Declare the make method which create an abstract term.
-# define TRM_IMPL_MAKE_DECL(Name, Base, Attributes, BaseArgs)   \
-    static                                                      \
-    term                                                        \
-    make(                                                       \
-      TRM_SEPARATE_COMMA(                                       \
-        TRM_MAP_ARRAY(TRM_CONST_ABSTRACT_COPY_ARG, Attributes)  \
-      )                                                         \
-    );
-
-// Declare all the attributes of the current class.
-# define TRM_ATTRIBUTE_DECLS(Name, Base, Attributes, BaseArgs)  \
-  public:                                                       \
-    TRM_APPLY(TRM_CONST_ABSTRACT_COPY_DECL, Attributes)
-
-
-// A class which provide static method (no virtual) to access the
-// implementation of the term.  This class is a wrapper over a pointer which
-// should derivate from the ATerm class.
-# define TRM_ABSTRACT_INTERFACE(Name, Base, Attributes, BaseArgs)       \
-  class A ## Name : public A ## Base                                    \
-  {                                                                     \
-    typedef A ## Base parent;                                           \
-                                                                        \
-  public:                                                               \
-    A ## Name (const A ## Name& t);                                     \
-    A ## Name ();                                                       \
-                                                                        \
-  protected:                                                            \
-    explicit A ## Name (const ATermImpl* ptr);                          \
-  };
-
-// Add the implementation class as a friend class to provide access to the
-// private constructor.  This reduce interaction with the implementation of
-// the terms.
-# define TRM_ABSTRACT_FINAL(Name, Base, Attributes, BaseArgs)   \
-  class A ## Name : public A ## Base                            \
-  {                                                             \
-    typedef A ## Base parent;                                   \
-                                                                \
-  public:                                                       \
-    A ## Name (const A ## Name& t);                             \
-    A ## Name ();                                               \
-                                                                \
-    const Name &                                                \
-    operator* () const;                                         \
-                                                                \
-    const Name*                                                 \
-    operator-> () const;                                        \
-                                                                \
-  protected:                                                    \
-    explicit A ## Name (const ATermImpl* ptr);                  \
-                                                                \
-    friend class Name;                                          \
-  };
-
-# define TRM_IMPL_INTERFACE(Name, Base, Attributes, BaseArgs)   \
-  class Name : public Base                                      \
-  {                                                             \
-    typedef Name this_type;                                     \
-    typedef Base parent;                                        \
-                                                                \
-  public:                                                       \
-    bool operator <(const Name& arg_rhs) const;                 \
-                                                                \
-  protected:                                                    \
-    TRM_IMPL_CTR_DECL(Name, Base, Attributes, BaseArgs)         \
-                                                                \
-  public:                                                       \
-    TRM_ATTRIBUTE_DECLS(Name, Base, Attributes, BaseArgs)       \
-  };
-
-# define TRM_IMPL_FINAL(Name, Base, Attributes, BaseArgs)       \
-  class Name : public Base                                      \
-  {                                                             \
-    typedef Name this_type;                                     \
-    typedef Base parent;                                        \
-  public:                                                       \
-    typedef A ## Name term;                                     \
-    typedef std::set<this_type> term_set_type;                  \
-                                                                \
-  public:                                                       \
-    TRM_IMPL_MAKE_DECL(Name, Base, Attributes, BaseArgs)        \
-    ATerm accept(ATermVisitor& v) const;                        \
-    bool operator <(const Name& arg_rhs) const;                 \
-                                                                \
-  protected:                                                    \
-    TRM_IMPL_CTR_DECL(Name, Base, Attributes, BaseArgs)         \
-                                                                \
-  public:                                                       \
-    TRM_ATTRIBUTE_DECLS(Name, Base, Attributes, BaseArgs)       \
-                                                                \
-  private:                                                      \
-                                                                \
-    static                                                      \
-    const ATermImpl*                                            \
-    get_identity(const this_type& t);                           \
-                                                                \
-    static                                                      \
-    term_set_type& set_instance();                              \
-  };
-
-
-  TRM_GRAMMAR_NODES(TRM_ABSTRACT_INTERFACE, TRM_ABSTRACT_FINAL)
-  TRM_GRAMMAR_NODES(TRM_IMPL_INTERFACE, TRM_IMPL_FINAL)
-
-# undef TRM_ABSTRACT_INTERFACE
-# undef TRM_ABSTRACT_FINAL
-# undef TRM_IMPL_INTERFACE
-# undef TRM_IMPL_FINAL
-
-# undef TRM_IMPL_CTR_DECL
-# undef TRM_ATTRIBUTE_DECLS
-# undef TRM_IMPL_MAKE_DECL
-  // ]
-
-
-# define TRM_GENERIC_MAKE_DECL(Name, Base, Attributes, BaseArgs)        \
-  Name :: term                                                          \
-  make ## Name(                                                         \
-    TRM_SEPARATE_COMMA(                                                 \
-      TRM_MAP_ARRAY(TRM_CONST_ABSTRACT_COPY_ARG, Attributes)            \
-    )                                                                   \
-  );
-
-  TRM_GRAMMAR_NODES(TRM_NIL_ARGS, TRM_GENERIC_MAKE_DECL)
-
-# undef TRM_GENERIC_MAKE_DECL
-}
-
-#endif

src/libterm/term_fwd.hh

@@ -1,28 +0,0 @@
-
-#ifndef _LIBTERM_TERM_FWD_HH
-# define _LIBTERM_TERM_FWD_HH
-
-# include "grammar.hh"
-
-// Types prefixed by an 'A' are the terms which should be manipulated by the
-// user.  The 'A' letter stands for Abstract.  Their content should be
-// limited to a pointer and they must not use any virtual keyword.  On the
-// other hand, the implementation of terms can declare more than one
-// attribute and use virtual keywords as long as they don't cost too much.
-
-namespace term
-{
-  class ATerm;
-  class ATermImpl; // should be renamed to TermImpl
-
-# define TRM_DECLARE(Name, Base, Attributes, BaseArgs)  \
-  class A ## Name;                                      \
-  class Name;
-
-  TRM_GRAMMAR_NODES(TRM_DECLARE, TRM_DECLARE)
-
-# undef TRM_DECLARE
-
-}
-
-#endif

src/libterm/term_impl.hh

@@ -1,269 +0,0 @@
-
-#ifndef _LIBTERM_TERM_HH
-# define _LIBTERM_TERM_HH
-
-# include "term.hh"
-# include "visitor.hh"
-
-namespace term
-{
-  inline
-  ATermImpl::ATermImpl()
-  {
-  }
-
-  inline
-  ATermImpl::~ATermImpl()
-  {
-  }
-
-
-  inline
-  ATerm::ATerm(const ATerm& t)
-    : ptr_(t.ptr_)
-  {
-  }
-
-  inline
-  ATerm::ATerm()
-    : ptr_(0)
-  {
-  }
-
-  inline
-  ATerm::ATerm(const ATermImpl* ptr)
-    : ptr_(ptr)
-  {
-  }
-
-  inline
-  ATerm
-  ATerm::accept(ATermVisitor& v) const
-  {
-    return ptr_->accept(v);
-  }
-
-  inline
-  bool
-  ATerm::operator== (const ATerm& rhs) const
-  {
-    return ptr_ == rhs.ptr_;
-  }
-
-  inline
-  bool
-  ATerm::operator!= (const ATerm& rhs) const
-  {
-    return ptr_ != rhs.ptr_;
-  }
-
-  inline
-  bool
-  ATerm::operator <(const ATerm& rhs) const
-  {
-    return ptr_ < rhs.ptr_;
-  }
-
-  inline
-  ATerm::operator bool()
-  {
-    return ptr_;
-  }
-
-  inline
-  const ATermImpl*
-  ATerm::get_ptr() const
-  {
-    return ptr_;
-  }
-
-
-  inline
-  bool
-  Term::operator <(const this_type&) const
-  {
-    return false;
-  }
-
-
-// define the contructors of the Abstract terms.  These constructors are
-// just delegating the work to the ATerm constructors.
-# define TRM_ABSTRACT_CTR(Name, Base, Attributes, BaseArgs)     \
-  inline                                                        \
-  A ## Name :: A ## Name (const A ## Name& t) :                 \
-    parent(t)                                                   \
-  {                                                             \
-  }                                                             \
-                                                                \
-  inline                                                        \
-  A ## Name :: A ## Name () :                                   \
-    parent()                                                    \
-  {                                                             \
-  }                                                             \
-                                                                \
-  inline                                                        \
-  A ## Name :: A ## Name (const ATermImpl* ptr) :               \
-    parent(ptr)                                                 \
-  {                                                             \
-  }
-
-  TRM_GRAMMAR_NODES(TRM_ABSTRACT_CTR, TRM_ABSTRACT_CTR)
-
-# undef TRM_ABSTRACT_CTR
-
-
-// Give acces to the implementation in order to get access to each field of
-// the term.
-# define TRM_ABSTRACT_FINAL_ACCESS(Name, Base, Attributes, BaseArgs)    \
-  inline                                                                \
-  const Name &                                                          \
-  A ## Name :: operator* () const                                       \
-  {                                                                     \
-    return *static_cast<const Name *>(ptr_);                            \
-  }                                                                     \
-                                                                        \
-  inline                                                                \
-  const Name*                                                           \
-  A ## Name :: operator-> () const                                      \
-  {                                                                     \
-    return static_cast<const Name *>(ptr_);                             \
-  }
-
-  TRM_GRAMMAR_NODES(TRM_NIL_ARGS, TRM_ABSTRACT_FINAL_ACCESS)
-
-# undef TRM_ABSTRACT_FINAL_ACCESS
-
-
-
-// Initialized all attributes of the current class and call the base class
-// with the expected arguments.
-# define TRM_IMPL_CTR(Name, Base, Attributes, BaseArgs)         \
-  inline                                                        \
-  Name :: Name (                                                \
-    TRM_SEPARATE_COMMA(                                         \
-      TRM_MAP_ARRAY(TRM_CONST_ABSTRACT_COPY_ARG, Attributes)    \
-    )                                                           \
-  )                                                             \
-    : TRM_SEPARATE_COMMA_SEQ(                                   \
-        TRM_MAP(TRM_INIT_ATTRIBUTES, Attributes)                \
-        ( parent(TRM_SEPARATE_COMMA(                            \
-            TRM_MAP_ARRAY(TRM_ARGUMENTS, BaseArgs)              \
-          ))                                                    \
-        )                                                       \
-      )                                                         \
-  {                                                             \
-  }
-
-  TRM_GRAMMAR_NODES(TRM_IMPL_CTR, TRM_IMPL_CTR)
-
-# undef TRM_IMPL_CTR
-
-
-// This operator is used for checking if the current ellement has not been
-// create before.  It compares each attributes of the current class and
-// delegate other attribute comparison to his parent class.
-# define TRM_IMPL_LESS(Name, Base, Attributes, BaseArgs)        \
-  inline                                                        \
-  bool                                                          \
-  Name :: operator <(const Name& arg_rhs) const                 \
-  {                                                             \
-    return TRM_APPLY(TRM_LESS_RHS_OR, Attributes)               \
-      parent::operator < (arg_rhs);                             \
-  }
-
-  TRM_GRAMMAR_NODES(TRM_IMPL_LESS, TRM_IMPL_LESS)
-
-# undef TRM_IMPL_LESS
-
-
-// This method is used to provide maximal sharing among node of the same
-// types.  This create the class with the private constructor and register
-// it if this is not already done.  It returns a term which has a small
-// memory impact (only a pointer) which refers to the created element.
-# define TRM_IMPL_FINAL_MAKE(Name, Base, Attributes, BaseArgs)  \
-  inline                                                        \
-  Name :: term                                                  \
-  Name :: make(                                                 \
-    TRM_SEPARATE_COMMA(                                         \
-      TRM_MAP_ARRAY(TRM_CONST_ABSTRACT_COPY_ARG, Attributes)    \
-    )                                                           \
-  )                                                             \
-  {                                                             \
-    return term(get_identity(this_type(                         \
-      TRM_SEPARATE_COMMA(                                       \
-        TRM_MAP_ARRAY(TRM_ARGUMENTS, Attributes)                \
-      )                                                         \
-    )));                                                        \
-  }
-
-  TRM_GRAMMAR_NODES(TRM_NIL_ARGS, TRM_IMPL_FINAL_MAKE)
-
-# undef TRM_IMPL_FINAL_MAKE
-
-
-// Handle visit from the ATermVisitor and forward them to the Abstract term
-// of the current term.
-# define TRM_IMPL_FINAL_VISIT(Name, Base, Attributes, BaseArgs)         \
-  inline                                                                \
-  ATerm                                                                 \
-  Name :: accept(ATermVisitor& v) const                                 \
-  {                                                                     \
-    return v.visit(term(this));                                         \
-  }
-
-  TRM_GRAMMAR_NODES(TRM_NIL_ARGS, TRM_IMPL_FINAL_VISIT)
-
-# undef TRM_IMPL_FINAL_VISIT
-
-
-// These functions are used to provide maximal sharing between the terms.
-// All terms are stored inside a set only the address the element contained
-// in the set is used.
-# define TRM_IMPL_FINAL_SHARED(Name, Base, Attributes, BaseArgs)        \
-  inline                                                                \
-  const ATermImpl*                                                      \
-  Name :: get_identity(const this_type& t)                              \
-  {                                                                     \
-    return static_cast<const ATermImpl*>(                               \
-      &*set_instance().insert(t).first                                  \
-    );                                                                  \
-  }                                                                     \
-                                                                        \
-  inline                                                                \
-  Name :: term_set_type&                                                \
-  Name :: set_instance()                                                \
-  {                                                                     \
-    static term_set_type set;                                           \
-    return set;                                                         \
-  }
-
-  TRM_GRAMMAR_NODES(TRM_NIL_ARGS, TRM_IMPL_FINAL_SHARED)
-
-# undef TRM_IMPL_FINAL_SHARED
-
-
-// Improve user experience by providing function to create Abstract terms
-// without any manipulation of the implementation class names.
-# define TRM_GENERIC_MAKE(Name, Base, Attributes, BaseArgs)     \
-  inline                                                        \
-  Name :: term                                                  \
-  make ## Name(                                                 \
-    TRM_SEPARATE_COMMA(                                         \
-      TRM_MAP_ARRAY(TRM_CONST_ABSTRACT_COPY_ARG, Attributes)    \
-    )                                                           \
-  )                                                             \
-  {                                                             \
-    return Name :: make(                                        \
-      TRM_SEPARATE_COMMA(                                       \
-        TRM_MAP_ARRAY(TRM_ARGUMENTS, Attributes)                \
-      )                                                         \
-    );                                                          \
-  }
-
-  TRM_GRAMMAR_NODES(TRM_NIL_ARGS, TRM_GENERIC_MAKE)
-
-# undef TRM_GENERIC_MAKE
-
-}
-
-#endif

src/libterm/test.cc

@@ -1,56 +0,0 @@
-
-#include <vector>
-#include <iostream>
-
-#define TRM_GRAMMAR_NODES(Interface, Final)                     \
-  Interface(Expr, Term, (0, ()), (0, ()))                       \
-  TRM_GRAMMAR_NODE_BINOP(Final, Plus)                           \
-  Final(Int, Expr, (1, (TRM_TYPE_COPY(int, value))), (0, ()))
-
-#include "term_impl.hh"
-#include "visitor_impl.hh"
-#undef TRM_GRAMMAR_NODES
-
-using namespace term;
-
-struct Eval : public ATermVisitor
-{
-  int run(const ATerm t)
-  {
-    return as<AInt>(t.accept(*this))->value;
-  }
-
-  ATerm visit(const APlus p)
-  {
-    return Int::make(run(p->lhs) + run(p->rhs));
-  }
-};
-
-#define CHECK(Cond, Msg)                        \
-  if (Cond)                                     \
-  {                                             \
-    good++;                                     \
-    std::cout << "Ok: " << Msg << std::endl;    \
-  }                                             \
-  else                                          \
-  {                                             \
-    std::cout << "Ko: " << Msg << std::endl;    \
-  }                                             \
-  tests++
-
-int main()
-{
-  unsigned good, tests;
-
-  using namespace term;
-  AInt a = makeInt(1);
-  AInt b = makeInt(2);
-  AInt c = makeInt(1);
-  Eval e;
-
-  CHECK(a == c, "Terms are shared.");
-  CHECK(!as<APlus>(a), "Bad convertion returns a zero ATerm.");
-  CHECK(as<AInt>(a) == a, "Good convertion returns the same ATerm.");
-  CHECK(e.run(makePlus(a, makePlus(b, c))) == 4, "Visitors are working.");
-  return tests - good;
-}

src/libterm/visitor.hh

@@ -1,23 +0,0 @@
-
-#ifndef _LIBTERM_VISITOR_CLASS_FWD_HH
-# define _LIBTERM_VISITOR_CLASS_FWD_HH
-
-# include "term.hh"
-
-namespace term
-{
-  // base class for visitor implementation.
-  class ATermVisitor
-  {
-  public:
-# define TRM_VISITOR(Name, Base, Attributes, BaseArgs)  \
-    virtual ATerm visit(const A ## Name);
-
-    TRM_VISITOR(Term, TRM_NIL, TRM_NIL, TRM_NIL)
-    TRM_GRAMMAR_NODES(TRM_VISITOR, TRM_VISITOR)
-# undef TRM_VISITOR
-  };
-}
-
-#endif
-

src/libterm/visitor_fwd.hh

@@ -1,19 +0,0 @@
-
-#ifndef _LIBTERM_VISITOR_FWD_HH
-# define _LIBTERM_VISITOR_FWD_HH
-
-# include "term_fwd.hh"
-
-// Types prefixed by an 'A' are the terms which should be manipulated by the
-// user.  The 'A' letter stands for Abstract.  Their content should be
-// limited to a pointer and they must not use any virtual keyword.  On the
-// other hand, the implementation of terms can declare more than one
-// attribute and use virtual keywords as long as they don't cost too much.
-
-namespace term
-{
-  // base class for visitor implementation.
-  class ATermVisitor;
-}
-
-#endif

src/libterm/visitor_impl.hh

@@ -1,57 +0,0 @@
-
-#ifndef _LIBTERM_VISITOR_HH
-# define _LIBTERM_VISITOR_HH
-
-# include "visitor.hh"
-# include "term.hh"
-
-namespace term
-{
-
-// definition of the ATermVisitor visit functions.
-# define TRM_VISITOR(Name, Base, Attributes, BaseArgs)  \
-  ATerm                                                 \
-  ATermVisitor::visit(const A ## Name t) {              \
-    return t;                                           \
-  }
-
-  TRM_VISITOR(Term, TRM_NIL, TRM_NIL, TRM_NIL)
-  TRM_GRAMMAR_NODES(TRM_VISITOR, TRM_VISITOR)
-
-# undef TRM_VISITOR
-
-  namespace impl
-  {
-    template <typename T>
-    class asVisitor : ATermVisitor
-    {
-    public:
-      asVisitor(ATerm t)
-        : res()
-      {
-        t.accept(*this);
-      }
-
-      ATerm visit(const T t)
-      {
-        return res = t;
-      }
-
-    public:
-      T res;
-    };
-  }
-
-  // This function will return a zero ATerm if the element does not have the
-  // expected type.
-  template <typename T>
-  inline
-  T as(ATerm t)
-  {
-    impl::asVisitor<T> v(t);
-    return v.res;
-  }
-
-}
-
-#endif

src/libutil/Makefile.am

@@ -1,16 +1,16 @@
 pkglib_LTLIBRARIES = libutil.la
 
 libutil_la_SOURCES = util.cc hash.cc serialise.cc \
-  archive.cc aterm.cc aterm-map.cc xml-writer.cc
+  archive.cc xml-writer.cc
 
-libutil_la_LIBADD = ../boost/format/libformat.la
+libutil_la_LIBADD = ../boost/format/libformat.la ${aterm_lib} ${sqlite_lib}
 
 pkginclude_HEADERS = util.hh hash.hh serialise.hh \
-  archive.hh aterm.hh aterm-map.hh xml-writer.hh types.hh
+  archive.hh xml-writer.hh types.hh
 
 if !HAVE_OPENSSL
 libutil_la_SOURCES += \
  md5.c md5.h sha1.c sha1.h sha256.c sha256.h md32_common.h
 endif
 
-AM_CXXFLAGS = -Wall -I$(srcdir)/.. ${aterm_include}
+AM_CXXFLAGS = -Wall -I$(srcdir)/..

src/libutil/archive.cc

@@ -181,8 +181,6 @@ static void parseContents(ParseSink & sink, Source & source, const Path & path)
         left -= n;
     }
 
-    sink.finalizeContents(size);
-
     readPadding(size, source);
 }
 
@@ -317,12 +315,6 @@ struct RestoreSink : ParseSink
         writeFull(fd, data, len);
     }
 
-    void finalizeContents(unsigned long long size)
-    {
-        errno = ftruncate(fd, size);
-        if (errno) throw SysError(format("truncating file to its allocated length of %1% bytes") % size);
-    }
-
     void createSymlink(const Path & path, const string & target)
     {
         Path p = dstPath + path;