jake%bugzilla.org 041e66b04a The source files for the Bugzilla Guide have long been using the XML version of DocBook but still residing in the sgml/ directory with an extension of .sgml.
In an effort to maintain CVS history, the raw files were copied on the CVS server to the xml/ directory and renamed to have .xml for the extension; any checkins before this one did have the .sgml extension.


git-svn-id: svn://10.0.0.236/trunk@249206 18797224-902f-48f8-a5cc-f745e15eee43
2008-04-04 06:49:38 +00:00

1616 lines
56 KiB
XML

<!-- <!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.1.2//EN"> -->
<chapter id="installation" xreflabel="Bugzilla Installation">
<title>Installation</title>
<section id="stepbystep" xreflabel="Bugzilla Installation Step-by-step">
<title>Step-by-step Install</title>
<section id="intstall-into">
<title>Introduction</title>
<para>Bugzilla has been successfully installed under Solaris, Linux,
and Win32. Win32 is not yet officially supported, but many people
have got it working fine.
Please see
<xref linkend="os-win32" />
for further advice on getting Bugzilla to work on Microsoft
Windows.</para>
</section>
<section id="install-package-list">
<title>Package List</title>
<note>
<para> If you are running the very most recent
version of Perl and MySQL (both the executables and development
libraries) on your system, you can skip these manual installation
steps for the Perl modules by using Bundle::Bugzilla; see
<xref linkend="bundlebugzilla" />.
</para>
</note>
<para>The software packages necessary for the proper running of
Bugzilla (with download links) are:
<orderedlist>
<listitem>
<para>
<ulink url="http://www.mysql.com/">MySQL database server</ulink>
(&min-mysql-ver; or greater)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.perl.org">Perl</ulink>
(&min-perl-ver;, 5.6.1 is recommended if you wish to
use Bundle::Bugzilla)
</para>
</listitem>
<listitem>
<para>Perl Modules (minimum version):
<orderedlist>
<listitem>
<para>
<ulink url="http://www.template-toolkit.org">Template</ulink>
(v&min-template-ver;)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.perldoc.com/perl5.6/lib/File/Temp.html">
File::Temp</ulink>
(&min-file-temp-ver;) (Prerequisite for Template)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.cpan.org/modules/by-module/AppConfig/">AppConfig
</ulink>
(&min-appconfig-ver;)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.cpan.org/authors/id/MUIR/modules/Text-Tabs%2BWrap-2001.0131.tar.gz">Text::Wrap</ulink>
(&min-text-wrap-ver;)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://search.cpan.org/search?dist=File-Spec">File::Spec
</ulink>
(&min-file-spec-ver;)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.cpan.org/modules/by-module/Data/">Data::Dumper
</ulink>
(&min-data-dumper-ver;)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.cpan.org/modules/by-module/Mysql/">DBD::mysql
</ulink>
(&min-dbd-mysql-ver;)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.cpan.org/modules/by-module/DBI/">DBI</ulink>
(&min-dbi-ver;)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.cpan.org/modules/by-module/Date/">Date::Parse
</ulink>
(&min-date-format-ver;)
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.cpan.org/modules/by-module/CGI/">CGI
</ulink>
(&min-cgi-ver;)
</para>
</listitem>
</orderedlist>
and, optionally:
<orderedlist>
<listitem>
<para>
<ulink url="http://www.cpan.org/modules/by-module/GD/">GD</ulink>
(&min-gd-ver;) for bug charting
</para>
</listitem>
<listitem>
<para>
GD::Graph
(&min-gd-graph-ver;) for bug charting
</para>
</listitem>
<listitem>
<para>
GD::Text::Align
(&min-gd-text-align-ver;) for bug charting
</para>
</listitem>
<listitem>
<para>
<ulink url="http://www.cpan.org/modules/by-module/Chart/">Chart::Base
</ulink>
(&min-chart-base-ver;) for bug charting
</para>
</listitem>
<listitem>
<para>
XML::Parser
(&min-xml-parser-ver;) for the XML interface
</para>
</listitem>
<listitem>
<para>
MIME::Parser
(&min-mime-parser-ver;) for the email interface
</para>
</listitem>
</orderedlist>
</para>
</listitem>
<listitem>
<para>
The web server of your choice.
<ulink url="http://www.apache.org/">Apache</ulink>
is highly recommended.
</para>
</listitem>
</orderedlist>
<warning>
<para>It is a good idea, while installing Bugzilla, to ensure that there
is some kind of firewall between you and the rest of the Internet,
because your machine may be insecure for periods during the install.
Many
installation steps require an active Internet connection to complete,
but you must take care to ensure that at no point is your machine
vulnerable to an attack.</para>
</warning>
</para>
</section>
<section id="install-mysql">
<title>MySQL</title>
<para>Visit the MySQL homepage at
<ulink url="http://www.mysql.com">www.mysql.com</ulink>
to grab and install the latest stable release of the server.
</para>
<note>
<para> Many of the binary
versions of MySQL store their data files in
<filename>/var</filename>.
On some Unix systems, this is part of a smaller root partition,
and may not have room for your bug database. You can set the data
directory as an option to <filename>configure</filename>
if you build MySQL from source yourself.</para>
</note>
<para>If you install from something other than an RPM or Debian
package, you will need to add <filename>mysqld</filename>
to your init scripts so the server daemon will come back up whenever
your machine reboots. Further discussion of UNIX init sequences are
beyond the scope of this guide.
</para>
<para>Change your init script to start
<filename>mysqld</filename>
with the ability to accept large packets. By default,
<filename>mysqld</filename>
only accepts packets up to 64K long. This limits the size of
attachments you may put on bugs. If you add
<option>-O max_allowed_packet=1M</option>
to the command that starts
<filename>mysqld</filename>
(or <filename>safe_mysqld</filename>),
then you will be able to have attachments up to about 1 megabyte.
There is a Bugzilla parameter for maximum attachment size;
you should configure it to match the value you choose here.</para>
<para>If you plan on running Bugzilla and MySQL on the same machine,
consider using the
<option>--skip-networking</option>
option in the init script. This enhances security by preventing
network access to MySQL.</para>
</section>
<section id="install-perl">
<title>Perl</title>
<para>Any machine that doesn't have Perl on it is a sad machine indeed.
Perl can be got in source form from
<ulink url="http://www.perl.com">perl.com</ulink> for the rare
*nix systems which don't have it.
Although Bugzilla runs with perl &min-perl-ver;,
it's a good idea to be up to the very latest version
if you can when running Bugzilla. As of this writing, that is Perl
version &newest-perl-ver;.</para>
<tip id="bundlebugzilla"
xreflabel="Using Bundle::Bugzilla instead of manually installing Perl modules">
<para>You can skip the following Perl module installation steps by
installing
<productname>Bundle::Bugzilla</productname>
from
<glossterm linkend="gloss-cpan">CPAN</glossterm>,
which installs all required modules for you.</para>
<para>
<computeroutput>
<prompt>bash#</prompt>
<command>perl -MCPAN -e 'install "Bundle::Bugzilla"'</command>
</computeroutput>
</para>
<para>Bundle::Bugzilla doesn't include GD, Chart::Base, or
MIME::Parser, which are not essential to a basic Bugzilla install. If
installing this bundle fails, you should install each module
individually to isolate the problem.</para>
</tip>
</section>
<section id="perl-modules">
<title>Perl Modules</title>
<para>
All Perl modules can be found on the
<ulink url="http://www.cpan.org">Comprehensive Perl
Archive Network</ulink> (CPAN). The
CPAN servers have a real tendency to bog down, so please use mirrors.
</para>
<para>Quality, general Perl module installation instructions can be
found on the CPAN website, but the easy thing to do is to just use the
CPAN shell which does all the hard work for you.
To use the CPAN shell to install a module:
</para>
<para>
<computeroutput>
<prompt>bash#</prompt>
<command>perl -MCPAN -e 'install "&lt;modulename&gt;"'</command>
</computeroutput>
</para>
<para>
To do it the hard way:
</para>
<para>Untar the module tarball -- it should create its own
directory</para>
<para>CD to the directory just created, and enter the following
commands:
<orderedlist>
<listitem>
<para>
<computeroutput>
<prompt>bash#</prompt>
<command>perl Makefile.PL</command>
</computeroutput>
</para>
</listitem>
<listitem>
<para>
<computeroutput>
<prompt>bash#</prompt>
<command>make</command>
</computeroutput>
</para>
</listitem>
<listitem>
<para>
<computeroutput>
<prompt>bash#</prompt>
<command>make test</command>
</computeroutput>
</para>
</listitem>
<listitem>
<para>
<computeroutput>
<prompt>bash#</prompt>
<command>make install</command>
</computeroutput>
</para>
</listitem>
</orderedlist>
</para>
<warning>
<para>Many people complain that Perl modules will not install for
them. Most times, the error messages complain that they are missing a
file in
<quote>@INC</quote>.
Virtually every time, this error is due to permissions being set too
restrictively for you to compile Perl modules or not having the
necessary Perl development libraries installed on your system.
Consult your local UNIX systems administrator for help solving these
permissions issues; if you
<emphasis>are</emphasis>
the local UNIX sysadmin, please consult the newsgroup/mailing list
for further assistance or hire someone to help you out.</para>
</warning>
<section>
<title>DBI</title>
<para>The DBI module is a generic Perl module used the
MySQL-related modules. As long as your Perl installation was done
correctly the DBI module should be a breeze. It's a mixed Perl/C
module, but Perl's MakeMaker system simplifies the C compilation
greatly.</para>
</section>
<section>
<title>Data::Dumper</title>
<para>The Data::Dumper module provides data structure persistence for
Perl (similar to Java's serialization). It comes with later
sub-releases of Perl 5.004, but a re-installation just to be sure it's
available won't hurt anything.</para>
</section>
<section>
<title>MySQL-related modules</title>
<para>The Perl/MySQL interface requires a few mutually-dependent Perl
modules. These modules are grouped together into the the
Msql-Mysql-modules package.</para>
<para>The MakeMaker process will ask you a few questions about the
desired compilation target and your MySQL installation. For most of the
questions the provided default will be adequate, but when asked if your
desired target is the MySQL or mSQL packages, you should
select the MySQL related ones. Later you will be asked if you wish to
provide backwards compatibility with the older MySQL packages; you
should answer YES to this question. The default is NO.</para>
<para>A host of 'localhost' should be fine and a testing user of 'test'
with a null password should find itself with sufficient access to run
tests on the 'test' database which MySQL created upon installation.
</para>
</section>
<section>
<title>TimeDate modules</title>
<para>Many of the more common date/time/calendar related Perl modules
have been grouped into a bundle similar to the MySQL modules bundle.
This bundle is stored on the CPAN under the name TimeDate.
The component module we're most interested in is the Date::Format
module, but installing all of them is probably a good idea anyway.
</para>
</section>
<section>
<title>GD (optional)</title>
<para>The GD library was written by Thomas Boutell a long while ago to
programatically generate images in C. Since then it's become the
defacto standard for programmatic image construction. The Perl bindings
to it found in the GD library are used on millions of web pages to
generate graphs on the fly. That's what Bugzilla will be using it for
so you must install it if you want any of the graphing to work.</para>
<note>
<para>The Perl GD library requires some other libraries that may or
may not be installed on your system, including
<classname>libpng</classname>
and
<classname>libgd</classname>.
The full requirements are listed in the Perl GD library README.
If compiling GD fails, it's probably because you're
missing a required library.</para>
</note>
</section>
<section>
<title>Chart::Base (optional)</title>
<para>The Chart module provides Bugzilla with on-the-fly charting
abilities. It can be installed in the usual fashion after it has been
fetched from CPAN.
Note that earlier versions that 0.99c used GIFs, which are no longer
supported by the latest versions of GD.</para>
</section>
<section>
<title>Template Toolkit</title>
<para>When you install Template Toolkit, you'll get asked various
questions about features to enable. The defaults are fine, except
that it is recommended you use the high speed XS Stash of the Template
Toolkit, in order to achieve best performance.
</para>
</section>
</section>
<section id="sbs-http">
<title>HTTP Server</title>
<para>You have freedom of choice here, pretty much any web server that
is capable of running <glossterm linkend="gloss-cgi">CGI</glossterm>
scripts will work. <xref linkend="http"/> has more information about
configuring web servers to work with Bugzilla.
</para>
<note>
<para>We strongly recommend Apache as the web server to use. The
Bugzilla Guide installation instructions, in general, assume you are
using Apache. If you have got Bugzilla working using another webserver,
please share your experiences with us.</para>
</note>
</section>
<section>
<title>Bugzilla</title>
<para>You should untar the Bugzilla files into a directory that you're
willing to make writable by the default web server user (probably
<quote>nobody</quote>).
You may decide to put the files in the main web space for your
web server or perhaps in
<filename>/usr/local</filename>
with a symbolic link in the web space that points to the Bugzilla
directory.</para>
<tip>
<para>If you symlink the bugzilla directory into your Apache's HTML
hierarchy, you may receive
<errorname>Forbidden</errorname>
errors unless you add the
<quote>FollowSymLinks</quote>
directive to the &lt;Directory&gt; entry for the HTML root
in httpd.conf.</para>
</tip>
<para>Once all the files are in a web accessible directory, make that
directory writable by your webserver's user. This is a temporary step
until you run the post-install
<filename>checksetup.pl</filename>
script, which locks down your installation.</para>
</section>
<section>
<title>Setting Up the MySQL Database</title>
<para>After you've gotten all the software installed and working you're
ready to start preparing the database for its life as the back end to
a high quality bug tracker.</para>
<para>First, you'll want to fix MySQL permissions to allow access from
Bugzilla. For the purpose of this Installation section, the Bugzilla
username will be
<quote>bugs</quote>, and will have minimal permissions.
</para>
<para>Begin by giving the MySQL root user a password. MySQL passwords are limited
to 16 characters.
<simplelist>
<member>
<computeroutput>
<prompt>bash#</prompt>
<command>mysql -u root mysql</command>
</computeroutput>
</member>
<member>
<computeroutput>
<prompt>mysql&gt;</prompt>
<command>UPDATE user SET Password=PASSWORD('&lt;new_password'&gt;)
WHERE user='root';</command>
</computeroutput>
</member>
<member>
<computeroutput>
<prompt>mysql&gt;</prompt>
<command>FLUSH PRIVILEGES;</command>
</computeroutput>
</member>
</simplelist>
From this point on, if you need to access MySQL as the MySQL root user,
you will need to use
<command>mysql -u root -p</command>
and enter &lt;new_password&gt;. Remember that MySQL user names have
nothing to do with Unix user names (login names).</para>
<para>Next, we use an SQL <command>GRANT</command> command to create a
<quote>bugs</quote>
user, and grant sufficient permissions for checksetup.pl, which we'll
use later, to work its magic. This also restricts the
<quote>bugs</quote>
user to operations within a database called
<quote>bugs</quote>, and only allows the account to connect from
<quote>localhost</quote>.
Modify it to reflect your setup if you will be connecting from
another machine or as a different user.</para>
<para>Remember to set &lt;bugs_password&gt; to some unique password.
<simplelist>
<member>
<computeroutput>
<prompt>mysql&gt;</prompt>
<command>GRANT SELECT,INSERT,UPDATE,DELETE,INDEX,
ALTER,CREATE,DROP,REFERENCES ON bugs.* TO bugs@localhost
IDENTIFIED BY '&lt;bugs_password&gt;';</command>
</computeroutput>
</member>
<member>
<computeroutput>
<prompt>mysql&gt;</prompt>
<command>FLUSH PRIVILEGES;</command>
</computeroutput>
</member>
</simplelist>
</para>
<note>
<para>If you are using MySQL 4, the bugs user also needs to be granted
the LOCK TABLES and CREATE TEMPORARY TABLES permissions.
</para>
</note>
</section>
<section>
<title>
<filename>checksetup.pl</filename>
</title>
<para>Next, run the magic checksetup.pl script. (Many thanks to
<ulink url="mailto:holgerschurig@nikocity.de">Holger Schurig </ulink>
for writing this script!)
This script is designed to make sure your MySQL database and other
configuration options are consistent with the Bugzilla CGI files.
It will make sure Bugzilla files and directories have reasonable
permissions, set up the
<filename>data</filename>
directory, and create all the MySQL tables.
<simplelist>
<member>
<computeroutput>
<prompt>bash#</prompt>
<command>./checksetup.pl</command>
</computeroutput>
</member>
</simplelist>
The first time you run it, it will create a file called
<filename>localconfig</filename>.</para>
<para>This file contains a variety of settings you may need to tweak
including how Bugzilla should connect to the MySQL database.</para>
<para>The connection settings include:
<orderedlist>
<listitem>
<para>server's host: just use
<quote>localhost</quote>
if the MySQL server is local</para>
</listitem>
<listitem>
<para>database name:
<quote>bugs</quote>
if you're following these directions</para>
</listitem>
<listitem>
<para>MySQL username:
<quote>bugs</quote>
if you're following these directions</para>
</listitem>
<listitem>
<para>Password for the
<quote>bugs</quote>
MySQL account; (&lt;bugs_password&gt;) above</para>
</listitem>
</orderedlist>
</para>
<para>Once you are happy with the settings,
<filename>su</filename> to the user
your web server runs as, and re-run
<filename>checksetup.pl</filename>. (Note: on some security-conscious
systems, you may need to change the login shell for the webserver
account before you can do this.)
On this second run, it will create the database and an administrator
account for which you will be prompted to provide information.</para>
<note>
<para>The checksetup.pl script is designed so that you can run it at
any time without causing harm. You should run it after any upgrade to
Bugzilla.</para>
</note>
</section>
<section>
<title>Configuring Bugzilla</title>
<para>
You should run through the parameters on the Edit Parameters page
(link in the footer) and set them all to appropriate values.
They key parameters are documented in <xref linkend="parameters" />.
</para>
</section>
</section>
<section id="extraconfig">
<title>Optional Additional Configuration</title>
<section>
<title>Dependency Charts</title>
<para>As well as the text-based dependency graphs, Bugzilla also
supports dependency graphing, using a package called 'dot'.
Exactly how this works is controlled by the 'webdotbase' parameter,
which can have one of three values:
</para>
<para>
<orderedlist>
<listitem>
<para>
A complete file path to the command 'dot' (part of
<ulink url="http://www.graphviz.org/">GraphViz</ulink>)
will generate the graphs locally
</para>
</listitem>
<listitem>
<para>
A URL prefix pointing to an installation of the webdot package will
generate the graphs remotely
</para>
</listitem>
<listitem>
<para>
A blank value will disable dependency graphing.
</para>
</listitem>
</orderedlist>
</para>
<para>So, to get this working, install
<ulink url="http://www.graphviz.org/">GraphViz</ulink>. If you
do that, you need to
<ulink url="http://httpd.apache.org/docs/mod/mod_imap.html">enable
server-side image maps</ulink> in Apache.
Alternatively, you could set up a webdot server, or use the AT&amp;T
public webdot server (the
default for the webdotbase param). Note that AT&amp;T's server won't work
if Bugzilla is only accessible using HARTS.
</para>
</section>
<section>
<title>Bug Graphs</title>
<para>As long as you installed the GD and Graph::Base Perl modules you
might as well turn on the nifty Bugzilla bug reporting graphs.</para>
<para>Add a cron entry like this to run
<filename>collectstats.pl</filename>
daily at 5 after midnight:
<simplelist>
<member>
<computeroutput>
<prompt>bash#</prompt>
<command>crontab -e</command>
</computeroutput>
</member>
<member>
<computeroutput>5 0 * * * cd &lt;your-bugzilla-directory&gt; ;
./collectstats.pl</computeroutput>
</member>
</simplelist>
</para>
<para>After two days have passed you'll be able to view bug graphs from
the Bug Reports page.</para>
</section>
<section>
<title>The Whining Cron</title>
<para>By now you have a fully functional Bugzilla, but what good are
bugs if they're not annoying? To help make those bugs more annoying you
can set up Bugzilla's automatic whining system to complain at engineers
which leave their bugs in the NEW state without triaging them.
</para>
<para>
This can be done by
adding the following command as a daily crontab entry (for help on that
see that crontab man page):
<simplelist>
<member>
<computeroutput>
<command>cd &lt;your-bugzilla-directory&gt; ;
./whineatnews.pl</command>
</computeroutput>
</member>
</simplelist>
</para>
<tip>
<para>Depending on your system, crontab may have several manpages.
The following command should lead you to the most useful page for
this purpose:
<programlisting>
man 5 crontab
</programlisting>
</para>
</tip>
</section>
<section id="bzldap">
<title>LDAP Authentication</title>
<para>
<warning>
<para>This information on using the LDAP
authentication options with Bugzilla is old, and the authors do
not know of anyone who has tested it. Approach with caution.
</para>
</warning>
</para>
<para>
The existing authentication
scheme for Bugzilla uses email addresses as the primary user ID, and a
password to authenticate that user. All places within Bugzilla where
you need to deal with user ID (e.g assigning a bug) use the email
address. The LDAP authentication builds on top of this scheme, rather
than replacing it. The initial log in is done with a username and
password for the LDAP directory. This then fetches the email address
from LDAP and authenticates seamlessly in the standard Bugzilla
authentication scheme using this email address. If an account for this
address already exists in your Bugzilla system, it will log in to that
account. If no account for that email address exists, one is created at
the time of login. (In this case, Bugzilla will attempt to use the
"displayName" or "cn" attribute to determine the user's full name.)
After authentication, all other user-related tasks are still handled by
email address, not LDAP username. You still assign bugs by email
address, query on users by email address, etc.
</para>
<para>Using LDAP for Bugzilla authentication requires the
Mozilla::LDAP (aka PerLDAP) Perl module. The
Mozilla::LDAP module in turn requires Netscape's Directory SDK for C.
After you have installed the SDK, then install the PerLDAP module.
Mozilla::LDAP and the Directory SDK for C are both
<ulink url="http://www.mozilla.org/directory/">available for
download</ulink> from mozilla.org.
</para>
<para>
Set the Param 'useLDAP' to "On" **only** if you will be using an LDAP
directory for
authentication. Be very careful when setting up this parameter; if you
set LDAP authentication, but do not have a valid LDAP directory set up,
you will not be able to log back in to Bugzilla once you log out. (If
this happens, you can get back in by manually editing the data/params
file, and setting useLDAP back to 0.)
</para>
<para>If using LDAP, you must set the
three additional parameters: Set LDAPserver to the name (and optionally
port) of your LDAP server. If no port is specified, it defaults to the
default port of 389. (e.g "ldap.mycompany.com" or
"ldap.mycompany.com:1234") Set LDAPBaseDN to the base DN for searching
for users in your LDAP directory. (e.g. "ou=People,o=MyCompany") uids
must be unique under the DN specified here. Set LDAPmailattribute to
the name of the attribute in your LDAP directory which contains the
primary email address. On most directory servers available, this is
"mail", but you may need to change this.
</para>
<para>You can also try using <ulink url="http://www.openldap.org/">
OpenLDAP</ulink> with Bugzilla, using any of a number of administration
tools. You should apply the patch attached this bug:
<ulink url="http://bugzilla.mozilla.org/show_bug.cgi?id=158630">
http://bugzilla.mozilla.org/show_bug.cgi?id=158630</ulink>, then set
the following object classes for your users:
<orderedlist>
<listitem><para>objectClass: person</para></listitem>
<listitem><para>objectClass: organizationalPerson</para></listitem>
<listitem><para>objectClass: inetOrgPerson</para></listitem>
<listitem><para>objectClass: top</para></listitem>
<listitem><para>objectClass: posixAccount</para></listitem>
<listitem><para>objectClass: shadowAccount</para></listitem>
</orderedlist>
Please note that this patch <emphasis>has not</emphasis> yet been
accepted by the Bugzilla team, and so you may need to do some
manual tweaking. That said, it looks like Net::LDAP is probably
the way to go in the future.
</para>
</section>
<section id="content-type"
xreflabel="Preventing untrusted Bugzilla content from executing malicious Javascript code">
<title>Preventing untrusted Bugzilla content from executing malicious
Javascript code</title>
<para>It is possible for a Bugzilla to execute malicious Javascript
code. Due to internationalization concerns, we are unable to
incorporate the code changes necessary to fulfill the CERT advisory
requirements mentioned in
<ulink
url="http://www.cet.org/tech_tips/malicious_code_mitigation.html/#3">
http://www.cet.org/tech_tips/malicious_code_mitigation.html/#3</ulink>.
Executing the following code snippet from a UNIX command shell will
rectify the problem if your Bugzilla installation is intended for an
English-speaking audience. As always, be sure your Bugzilla
installation has a good backup before making changes, and I recommend
you understand what the script is doing before executing it.</para>
<para>
<programlisting>
bash# perl -pi -e "s/Content-Type\: text\/html/Content-Type\: text\/html\; charset=ISO-8859-1/i" *.cgi *.pl
</programlisting>
</para>
<para>All this one-liner command does is search for all instances of
<quote>Content-type: text/html</quote>
and replaces it with
<quote>Content-Type: text/html; charset=ISO-8859-1</quote>
. This specification prevents possible Javascript attacks on the
browser, and is suggested for all English-speaking sites. For
non-English-speaking Bugzilla sites, I suggest changing
<quote>ISO-8859-1</quote>, above, to
<quote>UTF-8</quote>.</para>
<note>
<para>Using &lt;meta&gt; tags to set the charset is not
recommended, as there's a bug in Netscape 4.x which causes pages
marked up in this way to load twice. See
<ulink url="http://bugzilla.mozilla.org/show_bug.cgi?id=126266">bug
126266</ulink> for more information including progress toward making
bugzilla charset aware by default.
</para>
</note>
</section>
<section id="directoryindex" xreflabel="Modifying the Apache
DirectoryIndex parameter to use index.cgi">
<title>
<filename>directoryindex</filename> for the Bugzilla default page.
</title>
<para>You should modify the &lt;DirectoryIndex&gt; parameter for
the Apache virtual host running your Bugzilla installation to
allow <filename>index.cgi</filename> as the index page for a
directory, as well as the usual <filename>index.html</filename>,
<filename>index.htm</filename>, and so forth. </para>
</section>
<section id="mod_perl" xreflabel="Bugzilla and mod_perl">
<title>
Bugzilla and <filename>mod_perl</filename>
</title>
<para>Bugzilla is unsupported under mod_perl. Effort is underway
to make it work cleanly in a mod_perl environment, but it is
slow going.
</para>
</section>
<section id="mod-throttle"
xreflabel="Using mod_throttle to prevent Denial of Service attacks">
<title>
<filename>mod_throttle</filename>
and Security</title>
<para>It is possible for a user, by mistake or on purpose, to access
the database many times in a row which can result in very slow access
speeds for other users. If your Bugzilla installation is experiencing
this problem , you may install the Apache module
<filename>mod_throttle</filename>
which can limit connections by ip-address. You may download this module
at
<ulink url="http://www.snert.com/Software/Throttle/">
http://www.snert.com/Software/Throttle/</ulink>.
Follow the instructions to install into your Apache install.
<emphasis>This module only functions with the Apache web
server!</emphasis>
You may use the
<command>ThrottleClientIP</command>
command provided by this module to accomplish this goal. See the
<ulink url="http://www.snert.com/Software/Throttle/">Module
Instructions</ulink>
for more information.</para>
</section>
</section>
<section id="os-specific">
<title>OS Specific Installation Notes</title>
<para>Many aspects of the Bugzilla installation can be affected by the
the operating system you choose to install it on. Sometimes it can be made
easier and others more difficult. This section will attempt to help you
understand both the difficulties of running on specific operating systems
and the utilities available to make it easier.
</para>
<para>If you have anything to add or notes for an operating system not
covered, please file a bug in &bzg-bugs;.
</para>
<section id="os-win32">
<title>Microsoft Windows</title>
<para>Making Bugzilla work on windows is still a very painful processes.
The Bugzilla Team is working to make it easier, but that goal is not
considered a top priority. If you wish to run Bugzilla, we still
recommend doing so on a Unix based system such as GNU/Linux. As of this
writing, all members of the Bugzilla team and all known large installations
run on Unix based systems.
</para>
<para>If after hearing all that, you have enough pain tolerance to attempt
installing Bugzilla on Win32, here are some pointers.
<![%bz-devel;[
Because this is a development version of the guide, these instructions
are subject to change without notice. In fact, the Bugzilla Team hopes
they do as we would like to have Bugzilla resonabally close to "out of
the box" compatibility by the 2.18 release.
]]>
</para>
<section id="win32-perl">
<title>Win32 Perl</title>
<para>Perl for Windows can be obtained from <ulink
url="http://www.activestate.com/">ActiveState</ulink>. You should be
able to find a compiled binary at <ulink
url="http://aspn.activestate.com/ASPN/Downloads/ActivePerl/">http://aspn.activestate.com/ASPN/Downloads/ActivePerl/</ulink>.
</para>
</section>
<section id="win32-perl-modules">
<title>Perl Modules on Win32</title>
<para>Bugzilla on Windows requires the same perl modules found in
<xref linkend="install-package-list"/>. The main difference is that
windows uses <command>ppm</command> instead of CPAN.
</para>
<programlisting>
C:\perl&gt; <command>ppm &lt;module name&gt;</command>
</programlisting>
<note>
<para>The above syntax should work for all modules with the exception
of Template Toolkit. The <ulink
url="http://tt2.org/download.html#win32">Template Toolkit website</ulink>
suggests using the instructions on <ulink
url="http://openinteract.sourceforge.net/">OpenInteract's website</ulink>.
</para>
</note>
<tip>
<para>A complete list of modules that can be installed using ppm can
be found at <ulink url="http://www.activestate.com/PPMPackages/5.6plus">http://www.activestate.com/PPMPackages/5.6plus</ulink>.
</para>
</tip>
</section>
<section id="win32-code-changes">
<title>Code changes required to run on win32</title>
<para>Unfortunately, Bugzilla still doesn't run "out of the box" on
Windows. There is work in progress to make this easier, but until that
happens code will have to be modified. This section is an attempt to
list the required changes. It is an attempt to be all inclusive, but
there may be other changes required. If you find something is missing,
please file a bug in &bzg-bugs;.
</para>
<section id="win32-code-checksetup">
<title>Changes to <filename>checksetup.pl</filename></title>
<para>In <filename>checksetup.pl</filename>, the line reading:</para>
<programlisting>
my $mysql_binaries = `which mysql`;
</programlisting>
<para>to</para>
<programlisting>
my $mysql_binaries = "D:\\mysql\\bin\\mysql";
</programlisting>
<para>And you'll also need to change:</para>
<programlisting>
my $webservergid = getgrnam($my_webservergroup)
</programlisting>
<para>to</para>
<programlisting>
my $webservergid = '8'
</programlisting>
</section>
</section>
<section id="win32-http">
<title>Serving the web pages</title>
<para>As is the case on Unix based systems, any web server should be
able to handle Bugzilla; however, the Bugzilla Team still recommends
Apache whenever asked. No matter what web server you choose, be sure
to pay attention to the security notes in <xref linkend="security-access"/>.
More information on configuring specific web servers can be found in
<xref linkend="http"/>.
</para>
<note>
<para>If using Apache on windows, you can set the <ulink
url="http://httpd.apache.org/docs-2.0/mod/core.html#scriptinterpretersource">ScriptInterpreterSource</ulink>
directive in your Apache config, if you don't do this, you'll have
to modify the first line of every script to contain your path to
perl instead of <filename>/usr/bin/perl</filename>.
</para>
</note>
</section>
</section>
<section id="os-macosx">
<title><productname>Mac OS X</productname></title>
<!-- TODO: Clean me up... (Mac OS X) -->
<para>There are a lot of common libraries and utilities out there that
Apple did not include with Mac OS X, but which run perfectly well on it.
The GD library, which Bugzilla needs to do bug graphs, is one of
these.</para>
<para>The easiest way to get a lot of these is with a program called
Fink, which is similar in nature to the CPAN installer, but installs
common GNU utilities. Fink is available from
<ulink url="http://sourceforge.net/projects/fink/"/>.</para>
<para>Follow the instructions for setting up Fink. Once it's installed,
you'll want to run the following as root:
<command>fink install gd</command>
</para>
<para>It will prompt you for a number of dependencies, type 'y' and hit
enter to install all of the dependencies. Then watch it work.</para>
<para>To prevent creating conflicts with the software that Apple installs
by default, Fink creates its own directory tree at /sw where it installs
most of the software that it installs. This means your libraries and
headers for libgd will be at /sw/lib and /sw/include instead of /usr/lib
and /usr/local/include. Because of these changed locations for the
libraries, the Perl GD module will not install directly via CPAN, because it
looks for the specific paths instead of getting them from your
environment. But there's a way around that :-)</para>
<para>Instead of typing
<quote>install GD</quote>
at the
<prompt>cpan&gt;</prompt>
prompt, type
<command>look GD</command>.
This should go through the motions of downloading the latest version of
the GD module, then it will open a shell and drop you into the build
directory. Apply <ulink url="../xml/gd-makefile.patch">this patch</ulink>
to the Makefile.PL file (save the
patch into a file and use the command
<command>patch &lt; patchfile</command>.)
</para>
<para>Then, run these commands to finish the installation of the GD
module:
<simplelist>
<member>
<command>perl Makefile.PL</command>
</member>
<member>
<command>make</command>
</member>
<member>
<command>make test</command>
</member>
<member>
<command>make install</command>
</member>
<member>And don't forget to run
<command>exit</command>
to get back to CPAN.</member>
</simplelist>
</para>
</section>
<section id="os-mandrake">
<title>Linux-Mandrake 8.0</title>
<para>Linux-Mandrake 8.0 includes every required and optional library
for Bugzilla. The easiest way to install them is by using the
<command>urpmi</command> utility. If you follow these commands, you
should have everything you need for Bugzilla, and
<command>./checksetup.pl</command> should not complain about any
missing libraries. You may already have some of these installed.
</para>
<screen>
<prompt>bash#</prompt> <command>urpmi perl-mysql</command>
<prompt>bash#</prompt> <command>urpmi perl-chart</command>
<prompt>bash#</prompt> <command>urpmi perl-gd</command>
<prompt>bash#</prompt> <command>urpmi perl-MailTools</command> <co id="test-mailtools"/>
<prompt>bash#</prompt> <command>urpmi apache-modules</command>
</screen>
<calloutlist>
<callout arearefs="test-mailtools">
<para>for Bugzilla e-mail integration</para>
</callout>
</calloutlist>
</section>
</section>
<section id="http">
<title>HTTP Server Configuration</title>
<para>The Bugzilla Team recommends Apache when using Bugzilla, however, any web server
that can be configured to run <glossterm linkend="gloss-cgi">CGI</glossterm> scripts
should be able to handle Bugzilla. No matter what web server you choose, but
especially if you choose something other than Apache, you should be sure to read
<xref linkend="security-access"/>.
</para>
<para>The plan for this section is to eventually document the specifics of how to lock
down permissions on individual web servers.
</para>
<section id="http-apache">
<title>Apache <productname>httpd</productname></title>
<para>As mentioned above, the Bugzilla Team recommends Apache for use
with Bugzilla. You will have to make sure that Apache is properly
configured to run the Bugzilla CGI scripts. You also need to make sure
that the <filename>.htaccess</filename> files created by
<command>./checksetup.pl</command> (shown in <xref linkend="http-apache-htaccess"/>
for the curious) are allowed to override Apache's normal access
permissions or else important password information may be exposed to the
Internet.
</para>
<para>Many Apache installations are not configured to run scripts
anywhere but in the <filename class="directory">cgi-bin</filename>
directory; however, we recommend that Bugzilla not be installed in the
<filename class="directory">cgi-bin</filename>, otherwise the static
files such as images and <xref linkend="gloss-javascript"/>
will not work correctly. To allow scripts to run in the normal
web space, the following changes should be made to your
<filename>httpd.conf</filename> file.
</para>
<para>To allow files with a .cgi extension to be run, make sure the
following line exists and is uncommented:</para>
<programlisting>
AddHandler cgi-script .cgi
</programlisting>
<para>To allow <filename>.htaccess</filename> files to override
permissions and .cgi files to run in the Bugzilla directory, make sure
the following two lines are in a <computeroutput>Directory</computeroutput>
directive that applies to the Bugzilla directory on your system
(either the Bugzilla directory or one of its parents).
</para>
<programlisting>
Options +ExecCGI
AllowOverride Limit
</programlisting>
<note>
<para>For more information on Apache and its directives, see the
glossary entry on <xref linkend="gloss-apache"/>.
</para>
</note>
<example id="http-apache-htaccess">
<title><filename>.htaccess</filename> files for Apache</title>
<para><filename>$BUGZILLA_HOME/.htaccess</filename>
<programlisting><![CDATA[
# don't allow people to retrieve non-cgi executable files or our private data
<FilesMatch ^(.*\.pl|.*localconfig.*|runtests.sh)$>
deny from all
</FilesMatch>
<FilesMatch ^(localconfig.js|localconfig.rdf)$>
allow from all
</FilesMatch>
]]></programlisting>
</para>
<para><filename>$BUGZILLA_HOME/data/.htaccess</filename>
<programlisting><![CDATA[
# nothing in this directory is retrievable unless overriden by an .htaccess
# in a subdirectory; the only exception is duplicates.rdf, which is used by
# duplicates.xul and must be loadable over the web
deny from all
<Files duplicates.rdf>
allow from all
</Files>
]]></programlisting>
</para>
<para><filename>$BUGZILLA_HOME/data/webdot</filename>
<programlisting><![CDATA[
# Restrict access to .dot files to the public webdot server at research.att.com
# if research.att.com ever changed their IP, or if you use a different
# webdot server, you'll need to edit this
<FilesMatch ^[0-9]+\.dot$>
Allow from 192.20.225.10
Deny from all
</FilesMatch>
# Allow access by a local copy of 'dot' to .png, .gif, .jpg, and
# .map files
<FilesMatch ^[0-9]+\.(png|gif|jpg|map)$>
Allow from all
</FilesMatch>
# And no directory listings, either.
Deny from all
]]></programlisting>
</para>
<para><filename>$BUGZILLA_HOME/Bugzilla/.htaccess</filename>
<programlisting>
# nothing in this directory is retrievable unless overriden by an .htaccess
# in a subdirectory
deny from all
</programlisting>
</para>
<para><filename>$BUGZILLA_HOME/template/.htaccess</filename>
<programlisting>
# nothing in this directory is retrievable unless overriden by an .htaccess
# in a subdirectory
deny from all
</programlisting>
</para>
</example>
</section>
<section id="http-iis">
<title>Microsoft <productname>Internet Information Services</productname></title>
<para>If you need, or for some reason even want, to use Microsoft's
<productname>Internet Information Services</productname> or
<productname>Personal Web Server</productname> you should be able
to. You will need to configure them to know how to run CGI scripts,
however. This is described in Microsoft Knowledge Base article
<ulink url="http://support.microsoft.com/support/kb/articles/Q245/2/25.asp">Q245225 </ulink>
for <productname>Internet Information Services</productname> and
<ulink url="http://support.microsoft.com/support/kb/articles/Q231/9/98.asp">Q231998</ulink>
for <productname>Personal Web Server</productname>.
</para>
<para>Also, and this can't be stressed enough, make sure that files such as
<filename>localconfig</filename> and your <filename class="directory">data</filename>
directory are secured as described in <xref linkend="security-access"/>.
</para>
</section>
<section id="http-aol">
<title>AOL Server</title>
<para>Ben FrantzDale reported success using AOL Server with Bugzilla. He
reported his experience and what appears below is based on that.
</para>
<para>AOL Server will have to be configured to run
<glossterm linkend="gloss-cgi">CGI</glossterm> scripts, please consult
the documentation that came with your server for more information on
how to do this.
</para>
<para>Because AOL Server doesn't support <filename>.htaccess</filename>
files, you'll have to create a <glossterm linkend="gloss-tcl">TCL</glossterm>
script. You should create an <filename>aolserver/modules/tcl/filter.tcl</filename>
file (the filename shouldn't matter) with the following contents (change
<computeroutput>/bugzilla/</computeroutput> to the web-based path to
your Bugzilla installation):
</para>
<programlisting>
ns_register_filter preauth GET /bugzilla/localconfig filter_deny
ns_register_filter preauth GET /bugzilla/localconfig~ filter_deny
ns_register_filter preauth GET /bugzilla/\#localconfig\# filter_deny
ns_register_filter preauth GET /bugzilla/*.pl filter_deny
ns_register_filter preauth GET /bugzilla/syncshadowdb filter_deny
ns_register_filter preauth GET /bugzilla/runtests.sh filter_deny
ns_register_filter preauth GET /bugzilla/data/* filter_deny
ns_register_filter preauth GET /bugzilla/template/* filter_deny
proc filter_deny { why } {
ns_log Notice "filter_deny"
return "filter_return"
}
</programlisting>
<warning>
<para>This probably doesn't account for all possible editor backup
files so you may wish to add some additional variations of
<filename>localconfig</filename>. For more information, see
<ulink url="http://bugzilla.mozilla.org/show_bug.cgi?id=186383">bug
186383</ulink> or <ulink
url="http://online.securityfocus.com/bid/6501">Bugtraq ID 6501</ulink>.
</para>
</warning>
<note>
<para>If you are using webdot from research.att.com (the default
configuration for the <option>webdotbase</option> paramater), you
will need to allow access to <filename>data/webdot/*.dot</filename>
for the reasearch.att.com machine.
</para>
<para>If you are using a local installation of <ulink
url="http://www.graphviz.org">GraphViz</ulink>, you will need to allow
everybody to access <filename>*.png</filename>,
<filename>*.gif</filename>, <filename>*.jpg</filename>, and
<filename>*.map</filename> in the
<filename class="directory">data/webdot</filename> directory.
</para>
</note>
</section>
</section>
<section id="troubleshooting">
<title>Troubleshooting</title>
<para>This section gives solutions to common Bugzilla installation
problems.
</para>
<section>
<title>Bundle::Bugzilla makes me upgrade to Perl 5.6.1</title>
<para>
Try executing <command>perl -MCPAN -e 'install CPAN'</command>
and then continuing.
</para>
<para>
Certain older versions of the CPAN toolset were somewhat naive about how
to upgrade Perl modules. When a couple of modules got rolled into the core
Perl distribution for 5.6.1, CPAN thought that the best way to get those
modules up to date was to haul down the Perl distribution itself and
build it. Needless to say, this has caused headaches for just about
everybody. Upgrading to a newer version of CPAN with the
commandline above should fix things.
</para>
</section>
<section>
<title>DBD::Sponge::db prepare failed</title>
<para>
The following error message may appear due to a bug in DBD::mysql
(over which the Bugzilla team have no control):
</para>
<programlisting><![CDATA[ DBD::Sponge::db prepare failed: Cannot determine NUM_OF_FIELDS at D:/Perl/site/lib/DBD/mysql.pm line 248.
SV = NULL(0x0) at 0x20fc444
REFCNT = 1
FLAGS = (PADBUSY,PADMY)
]]></programlisting>
<para>
To fix this, go to
<filename>&lt;path-to-perl&gt;/lib/DBD/sponge.pm</filename>
in your Perl installation and replace
</para>
<programlisting><![CDATA[ my $numFields;
if ($attribs->{'NUM_OF_FIELDS'}) {
$numFields = $attribs->{'NUM_OF_FIELDS'};
} elsif ($attribs->{'NAME'}) {
$numFields = @{$attribs->{NAME}};
]]></programlisting>
<para>
by
</para>
<programlisting><![CDATA[ my $numFields;
if ($attribs->{'NUM_OF_FIELDS'}) {
$numFields = $attribs->{'NUM_OF_FIELDS'};
} elsif ($attribs->{'NAMES'}) {
$numFields = @{$attribs->{NAMES}};
]]></programlisting>
<para>
(note the S added to NAME.)
</para>
</section>
<section id="paranoid-security">
<title>cannot chdir(/var/spool/mqueue)</title>
<para>If you are installing Bugzilla on SuSE Linux, or some other
distributions with
<quote>paranoid</quote>
security options, it is possible that the checksetup.pl script may fail
with the error:
<programlisting><![CDATA[cannot chdir(/var/spool/mqueue): Permission denied
]]></programlisting>
</para>
<para>
This is because your
<filename>/var/spool/mqueue</filename>
directory has a mode of
<quote>drwx------</quote>. Type
<command>chmod 755
<filename>/var/spool/mqueue</filename>
</command>
as root to fix this problem.
</para>
</section>
<section id="trouble-filetemp">
<title>Your vendor has not defined Fcntl macro O_NOINHERIT</title>
<para>This is caused by a bug in the version of
<productname>File::Temp</productname> that is distributed with perl
5.6.0. Many minor variations of this error have been reported. Examples
can be found in <xref linkend="trouble-filetemp-errors"/>.
</para>
<figure id="trouble-filetemp-errors">
<title>Other File::Temp error messages</title>
<programlisting>
Your vendor has not defined Fcntl macro O_NOINHERIT, used
at /usr/lib/perl5/site_perl/5.6.0/File/Temp.pm line 208.
Your vendor has not defined Fcntl macro O_EXLOCK, used
at /usr/lib/perl5/site_perl/5.6.0/File/Temp.pm line 210.
Your vendor has not defined Fcntl macro O_TEMPORARY, used
at /usr/lib/perl5/site_perl/5.6.0/File/Temp.pm line 233.
</programlisting>
</figure>
<para>Numerous people have reported that upgrading to version 5.6.1
or higher solved the problem for them. A less involved fix is to apply
the patch in <xref linkend="trouble-filetemp-patch"/>. The patch is also
available as a <ulink url="../xml/filetemp.patch">patch file</ulink>.
</para>
<figure id="trouble-filetemp-patch">
<title>Patch for File::Temp in Perl 5.6.0</title>
<programlisting><![CDATA[
--- File/Temp.pm.orig Thu Feb 6 16:26:00 2003
+++ File/Temp.pm Thu Feb 6 16:26:23 2003
@@ -205,6 +205,7 @@
# eg CGI::Carp
local $SIG{__DIE__} = sub {};
local $SIG{__WARN__} = sub {};
+ local *CORE::GLOBAL::die = sub {};
$bit = &$func();
1;
};
@@ -226,6 +227,7 @@
# eg CGI::Carp
local $SIG{__DIE__} = sub {};
local $SIG{__WARN__} = sub {};
+ local *CORE::GLOBAL::die = sub {};
$bit = &$func();
1;
};
]]></programlisting>
</figure>
</section>
</section>
</chapter>
<!-- Keep this comment at the end of the file
Local variables:
mode: sgml
sgml-always-quote-attributes:t
sgml-auto-insert-required-elements:t
sgml-balanced-tag-edit:t
sgml-exposed-tags:nil
sgml-general-insert-case:lower
sgml-indent-data:t
sgml-indent-step:2
sgml-local-catalogs:nil
sgml-local-ecat-files:nil
sgml-minimize-attributes:nil
sgml-namecase-general:t
sgml-omittag:t
sgml-parent-document:("Bugzilla-Guide.xml" "book" "chapter")
sgml-shorttag:t
sgml-tag-region-if-active:t
End:
-->