The patch series was recreated by applying it to a CMake git checkout at
v3.10.1, and then rebasing it onto v3.11.0, followed by patch resync
with CMake 3.11.1
Patch 0004-Implement-Qt5-static-plugin-support.patch had to be
extensively redone, to support a refactoring in the Qt5 auto-gen support
in CMake 3.11.0. The rest of the patches are changed only due to
artifacts of this rebasing.
* cmake: Upgrade CMake package to 3.10.1
0004-Implement-Qt5-static-plugin-support.patch does not apply cleanly.
* cmake: Rebase the patch series onto 3.10.1
The patch series was applied using `git am` to a CMake v3.9.6 git
checkout, and then those commits were rebased onto CMake v3.10.1.
`git format-patch` was used to extract the patches again, as was
presumably originally done. In this case, I kept the existing numbers to
keep the patch readable, and avoid confusing file renames.
Only 0004-Implement-Qt5-static-plugin-support.patch needed manual
changes.
* cmake: CMake now unconditionally enables server mode
* cmake: Remove dead unit-test munging
This code appears to have been updating the paths in a CMake Unit Test
to match the live paths in MSYS2.
However, it was disabled when added (see #2633), and doesn't make sense
since the test is only of the parsing of the output format, not the
contents of the output.
mingw-w64-ncurses - 6.0 20170715 - Update to latest version
mingw-w64-libarchive - 3.3.2 - Update to latest version and update patch for bcrypt. The logic is a little better than what was in the source-code
mingw-w64-cmake - 3.9.0 - Update to latest version
* cmake and some updated dependencies
mingw-w64-cmake - 3.8.2 - Update to latest release version and remove 0002-bootstrap-Set-DEFAULT_CODEPAGE-in-cmake_cxx_flags_En.patch . That patch is not needed for bootstrap
mingw-w64-expat - 2.2.1 - Update to latest version, add testing as part of build process. Add 002-fix-tests.patch so those can work in MSYS environment. remove expat-2.2.0-CVE-2016-0718-regression.patch as that is in the source-code.
mingw-w64-lncruses - 6.0 - 20170624 - Update to latest version
* Fix various PKGBUILD errors (#2631)
These PKGBUILD files could no longer be openend by makepkg due to
stricter parsing in regard to variable types. This fixes the types.
* libpeas: Python 3 rebuild (#2627)
* mbedtls: Updated to 2.5.1 (#2629)
* python-sphinx: Fix dependencies (#2628)
* jasper: Added patch file to fix static library link error. (#2623)
To use the static library you need to define JAS_DLL to zero.
-D JAS_DLL=0
* expat: Update to 2.2.1 (#2625)
Includes various security fixes.
Drop patch applied upstream.
* Armadillo: update to 7.950.1 (#2624)
Fixes#2618
* glib2: Update to 2.52.3 (#2621)
* nodejs: update to 6.11.0 and include npm (#2622)
* OsgEarth and gdal improvments (#2620)
* osgearth : active osgEarthQt module
* gdal : change PROJ4 library name gdal try to dlopen libproj-9.dll but correct name is libproj-12.dll
* python3: Make multiprocessing.Queue work; Add some tests. Fixes#837 (#2619)
With non-mingw builds POSIX_SEMAPHORES_NOT_ENABLED is not enabled and
sem_unlink is a no-op. Change it to always be provided on Windows instead.
This adds a minimal test suite which makes it easier to catch regressions
in the future for common MSYS2 related bugs.
* vulkan: add sdk (#2590)
* glslang: update for vulkan sdk
* vulkan: add sdk
*vulkan-headers are included in sdk
*vulkan-docs are separate package because of versions difference
* vulcan: Fix PKGBUILD. Add conflicts
* Update PKGBUILD
Attempt to fix this build.
* Update checksums
perform conversion for more than one character in cm_codecvt
codecvt::do_out returned std::codecvt::error when called for more than 1 char
the Ninja (and NMake) generator use this codecvt and therefore failed when writing files
since this works in the official CMake binary release
i’d infer that msvc calls this function ? one char at a time ?
i’m not sure this patch is correct, but it seems to work
upstream bugreport https://gitlab.kitware.com/cmake/cmake/issues/16893
* mbedtls: Updated to 2.4.0 (#1868)
* mingw-w64-clang 3.9.0 - rev 2 delete files created by patch (#1867)
Delete files created by patches in case we are not deleting from srource
directory.
* jemalloc: Update to 4.2.1 (#1866)
* lua sqlite3 wrapper (#1861)
* Adds a TUI in GNU debugger, GDB. (#1869)
* nghttp2 - 1.16.0 Update to latest version (#1871)
Do not use --enable-lib-only for featureproofing.
* Readline (#1872)
* readline - 7.0 - Update to latest veersion - NOTE: Will probably break stuff - rebuild what depends upon it
Rekeyed patch for new version and ifdef'ed out chown calls if not
available
* Rebuild some packages that depend upon readline for good measure
* libssh2 - 1.8.0 - Upgrade to latest version (#1876)
* Adding library: flickcurl (#1875)
* Adding library:flickcurl
* Flickcurl: fixing dependency on gtk-doc
* Libarchive (#1878)
* libarchive : 3.2.2
Update to altest version
* libarchive: 3.2.2 - add back dependency on lz4
* Nsis (#1879)
* nsis 3.0 update to release version
* mingw-w64-nsis: 33.
Download should point to main to main site, not mirror
Use https when possible
* g-i-parser: fix race condition (#1881)
* Remove rmtree debugging patch
It is shadowing exceptions thrown from the main "try" block of
GDumpParser._execute_binary_get_tree().
There does seem to be a race condition, but it seems to be related to
subprocess.check_call() and what it does on Windows. At the point
xml.[blah].parse() is called, "dump.xml" is often absent!
* More tolerant rmtreeing
If shutil.rmtree() is failing due to some funky race condition that results
in files the walk doesn't know about being left, then don't worry about it
and continue.
This may leave uniquely named junk lying around in the build tree, but that
is by design not a problem with g-i-scanner. Just don't install the
tmp-introspectXXXXXX folder by accident.
* giscanner debugging: asserts and waits
Make absence of dump.xml or functions.txt fatal, but only after giving them
plenty of time time to appear.
Yes, this is horrible code. But it fixes broken builds almost completely on
my 64-bit Win7 with NTFS.
* Wait for XML to be parsed completely
It's possible that dump.xml might exist but be incompletely written. This
nasty sleep-till-it-parses loop covers that case.
Plus the usual logging so we can see if it's really needed ☺
* gobject-introspection: bump pkgrel
* Updated to stable release 190600_20161030 (#1880)
A fresh new stable release since January 30, 2014!
* go: update to 1.7.3 (#1882)
* Cmake and some prerequisits
mingw-w64-cmake - 3.7.0 - Update to latest version, note dependency upon
libuv
mingw-w64-jsoncpp - 1.7.1 - Update to latest version
mingw-w64-libarchive - 3.2.2 - Changed build to cmake and made patch to
enable Crypto-Next-Generation using bcrypt
mingw-w64-libuv - 1.10.0 - Update to latest version and include two
checkins for bug-fixes.
* mbedtls: Updated to 2.4.0 (#1868)
* mingw-w64-clang 3.9.0 - rev 2 delete files created by patch (#1867)
Delete files created by patches in case we are not deleting from srource
directory.
* jemalloc: Update to 4.2.1 (#1866)
* lua sqlite3 wrapper (#1861)
* Adds a TUI in GNU debugger, GDB. (#1869)
* nghttp2 - 1.16.0 Update to latest version (#1871)
Do not use --enable-lib-only for featureproofing.
* Readline (#1872)
* readline - 7.0 - Update to latest veersion - NOTE: Will probably break stuff - rebuild what depends upon it
Rekeyed patch for new version and ifdef'ed out chown calls if not
available
* Rebuild some packages that depend upon readline for good measure
* libssh2 - 1.8.0 - Upgrade to latest version (#1876)
* Adding library: flickcurl (#1875)
* Adding library:flickcurl
* Flickcurl: fixing dependency on gtk-doc
* Libarchive (#1878)
* libarchive : 3.2.2
Update to altest version
* libarchive: 3.2.2 - add back dependency on lz4
* Nsis (#1879)
* nsis 3.0 update to release version
* mingw-w64-nsis: 33.
Download should point to main to main site, not mirror
Use https when possible
* g-i-parser: fix race condition (#1881)
* Remove rmtree debugging patch
It is shadowing exceptions thrown from the main "try" block of
GDumpParser._execute_binary_get_tree().
There does seem to be a race condition, but it seems to be related to
subprocess.check_call() and what it does on Windows. At the point
xml.[blah].parse() is called, "dump.xml" is often absent!
* More tolerant rmtreeing
If shutil.rmtree() is failing due to some funky race condition that results
in files the walk doesn't know about being left, then don't worry about it
and continue.
This may leave uniquely named junk lying around in the build tree, but that
is by design not a problem with g-i-scanner. Just don't install the
tmp-introspectXXXXXX folder by accident.
* giscanner debugging: asserts and waits
Make absence of dump.xml or functions.txt fatal, but only after giving them
plenty of time time to appear.
Yes, this is horrible code. But it fixes broken builds almost completely on
my 64-bit Win7 with NTFS.
* Wait for XML to be parsed completely
It's possible that dump.xml might exist but be incompletely written. This
nasty sleep-till-it-parses loop covers that case.
Plus the usual logging so we can see if it's really needed ☺
* gobject-introspection: bump pkgrel
* Updated to stable release 190600_20161030 (#1880)
A fresh new stable release since January 30, 2014!
* go: update to 1.7.3 (#1882)
* cmake - 3.6.3 - Update to latest version
* upgrade some broken packages
* correct some broken urls/checksums
* use secure urls where possible
* update PKGBUILD templates
* remove line-ending whitespace
* minor typos/fixes
Remaining md5sums either didn't download or
didn't pass checksum tests.
