LDP/LDP/retired/Bugzilla-Guide/administration.xml

1530 lines
58 KiB
XML

<!-- <!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook V4.1//EN"> -->
<chapter id="administration">
<title>Administering Bugzilla</title>
<section id="parameters">
<title>Bugzilla Configuration</title>
<para>Bugzilla is configured by changing various parameters, accessed
from the "Edit parameters" link in the page footer. Here are
some of the key parameters on that page. You should run down this
list and set them appropriately after installing Bugzilla.</para>
<indexterm>
<primary>checklist</primary>
</indexterm>
<procedure>
<step>
<para>
<command>maintainer</command>:
The maintainer parameter is the email address of the person
responsible for maintaining this
Bugzilla installation. The address need not be that of a valid Bugzilla
account.</para>
</step>
<step>
<para>
<command>urlbase</command>:
This parameter defines the fully qualified domain name and web
server path to your Bugzilla installation.</para>
<para>For example, if your Bugzilla query page is
<filename>http://www.foo.com/bugzilla/query.cgi</filename>,
set your <quote>urlbase</quote>
to <filename>http://www.foo.com/bugzilla/</filename>.</para>
</step>
<step>
<para>
<command>usebuggroups</command>:
This dictates whether or not to implement group-based security for
Bugzilla. If set, Bugzilla bugs can have an associated 'group',
defining which users are allowed to see and edit the
bug.</para>
<para>Set "usebuggroups" to "on"
<emphasis>only</emphasis>
if you may wish to restrict access to particular bugs to certain
groups of users. I suggest leaving
this parameter <emphasis>off</emphasis>
while initially testing your Bugzilla.</para>
</step>
<step>
<para>
<command>usebuggroupsentry</command>:
Bugzilla Products can have a group associated with them, so that
certain users can only see bugs in certain products. When this parameter
is set to <quote>on</quote>, this places all newly-created bugs in the
group for their product immediately.</para>
</step>
<step>
<para>
<command>shadowdb</command>:
You run into an interesting problem when Bugzilla reaches a
high level of continuous activity. MySQL supports only table-level
write locking. What this means is that if someone needs to make a
change to a bug, they will lock the entire table until the operation
is complete. Locking for write also blocks reads until the write is
complete. The
<quote>shadowdb</quote>
parameter was designed to get around this limitation. While only a
single user is allowed to write to a table at a time, reads can
continue unimpeded on a read-only shadow copy of the database.
Although your database size will double, a shadow database can cause
an enormous performance improvement when implemented on extremely
high-traffic Bugzilla databases.</para>
<para>
As a guide, mozilla.org began needing
<quote>shadowdb</quote>
when they reached around 40,000 Bugzilla users with several hundred
Bugzilla bug changes and comments per day.</para>
<para>The value of the parameter defines the name of the
shadow bug database.
Set "shadowdb" to e.g. "bug_shadowdb" if you will be running a
*very* large installation of Bugzilla.
<note>
<para>Enabling "shadowdb" can adversely affect the stability of
your installation of Bugzilla. You should regularly check that your
database is in sync. It is often advisable to force a shadow
database sync nightly via
<quote>cron</quote>.
</para>
</note>
</para>
<para>If you use the "shadowdb" option, it is only natural that you
should turn the "queryagainstshadowdb" option on as well. Otherwise
you are replicating data into a shadow database for no reason!</para>
</step>
<step>
<para>
<command>shutdownhtml</command>:
If you need to shut down Bugzilla to perform administration, enter
some descriptive HTML here and anyone who tries to use Bugzilla will
receive a page to that effect. Obviously, editparams.cgi will
still be accessible so you can remove the HTML and re-enable Bugzilla.
:-)
</para>
</step>
<step>
<para>
<command>passwordmail</command>:
Every time a user creates an account, the text of
this parameter (with substitutions) is sent to the new user along with
their password message.</para>
<para>Add any text you wish to the "passwordmail" parameter box. For
instance, many people choose to use this box to give a quick training
blurb about how to use Bugzilla at your site.</para>
</step>
<step>
<para>
<command>useqacontact</command>:
This allows you to define an email address for each component, in
addition
to that of the default owner, who will be sent carbon copies of
incoming bugs.</para>
</step>
<step>
<para>
<command>usestatuswhiteboard</command>:
This defines whether you wish to have a free-form, overwritable field
associated with each bug. The advantage of the Status Whiteboard is
that it can be deleted or modified with ease, and provides an
easily-searchable field for indexing some bugs that have some trait
in common.
</para>
</step>
<step>
<para>
<command>whinedays</command>:
Set this to the number of days you want to let bugs go
in the NEW or REOPENED state before notifying people they have
untouched new bugs. If you do not plan to use this feature, simply do
not set up the whining cron job described in the installation
instructions, or set this value to "0" (never whine).</para>
</step>
<step>
<para>
<command>commenton*</command>:
All these
fields allow you to dictate what changes can pass without comment,
and which must have a comment from the person who changed them.
Often, administrators will allow users to add themselves to the CC
list, accept bugs, or change the Status Whiteboard without adding a
comment as to their reasons for the change, yet require that most
other changes come with an explanation.</para>
<para>Set the "commenton" options according to your site policy. It
is a wise idea to require comments when users resolve, reassign, or
reopen bugs at the very least.
<note>
<para>It is generally far better to require a developer comment
when resolving bugs than not. Few things are more annoying to bug
database users than having a developer mark a bug "fixed" without
any comment as to what the fix was (or even that it was truly
fixed!)</para>
</note>
</para>
</step>
<step>
<para>
<command>supportwatchers</command>:
Turning on this option allows users to ask to receive copies of
all a particular other user's bug email. This is, of
course, subject to the groupset restrictions on the bug; if the
<quote>watcher</quote>
would not normally be allowed to view a bug, the watcher cannot get
around the system by setting herself up to watch the bugs of someone
with bugs outside her privileges. They would still only receive email
updates for those bugs she could normally view.</para>
</step>
</procedure>
</section>
<section id="useradmin">
<title>User Administration</title>
<section id="defaultuser">
<title>Creating the Default User</title>
<para>When you first run checksetup.pl after installing Bugzilla, it
will prompt you for the administrative username (email address) and
password for this "super user". If for some reason you delete
the "super user" account, re-running checksetup.pl will again prompt
you for this username and password.</para>
<tip>
<para>If you wish to add more administrative users, you must use the
MySQL interface. Run "mysql" from the command line, and use these
commands:
<simplelist>
<member>
<prompt>mysql&gt;</prompt>
<command>use bugs;</command>
</member>
<member>
<prompt>mysql&gt;</prompt>
<command>
update profiles set groupset=0x7ffffffffffffff where login_name =
"(user's login name)";
</command>
</member>
</simplelist>
</para>
<para>Yes, that is
<emphasis>fourteen</emphasis>
<quote>f</quote>
's. A whole lot of f-ing going on if you want to create a new
administator.</para>
</tip>
</section>
<section id="manageusers">
<title>Managing Other Users</title>
<section id="createnewusers">
<title>Creating new users</title>
<para>Your users can create their own user accounts by clicking the
"New Account" link at the bottom of each page (assuming they
aren't logged in as someone else already.) However, should you
desire to create user accounts ahead of time, here is how you do
it.</para>
<orderedlist>
<listitem>
<para>After logging in, click the "Users" link at the footer of
the query page, and then click "Add a new user".</para>
</listitem>
<listitem>
<para>Fill out the form presented. This page is self-explanatory.
When done, click "Submit".</para>
<note>
<para>Adding a user this way will
<emphasis>not</emphasis>
send an email informing them of their username and password.
While useful for creating dummy accounts (watchers which
shuttle mail to another system, for instance, or email
addresses which are a mailing list), in general it is
preferable to log out and use the
<quote>New Account</quote>
button to create users, as it will pre-populate all the
required fields and also notify the user of her account name
and password.</para>
</note>
</listitem>
</orderedlist>
</section>
<section id="modifyusers">
<title>Modifying Users</title>
<para>To see a specific user, search for their login name
in the box provided on the "Edit Users" page. To see all users,
leave the box blank.</para>
<para>You can search in different ways the listbox to the right
of the text entry box. You can match by
case-insensitive substring (the default),
regular expression, or a
<emphasis>reverse</emphasis>
regular expression match, which finds every user name which does NOT
match the regular expression. (Please see
the <command>man regexp</command>
manual page for details on regular expression syntax.)
</para>
<para>Once you have found your user, you can change the following
fields:</para>
<itemizedlist>
<listitem>
<para>
<emphasis>Login Name</emphasis>:
This is generally the user's full email address. However, if you
have are using the emailsuffix Param, this may just be the user's
login name. Note that users can now change their login names
themselves (to any valid email address.)
</para>
</listitem>
<listitem>
<para>
<emphasis>Real Name</emphasis>: The user's real name. Note that
Bugzilla does not require this to create an account.</para>
</listitem>
<listitem>
<para>
<emphasis>Password</emphasis>:
You can change the user's password here. Users can automatically
request a new password, so you shouldn't need to do this often.
If you want to disable an account, see Disable Text below.
</para>
</listitem>
<listitem>
<para>
<emphasis>Disable Text</emphasis>:
If you type anything in this box, including just a space, the
user is prevented from logging in, or making any changes to
bugs via the web interface.
The HTML you type in this box is presented to the user when
they attempt to perform these actions, and should explain
why the account was disabled.
<warning>
<para>Don't disable the administrator account!</para>
</warning>
<note>
<para>The user can still submit bugs via
the e-mail gateway, if you set it up, even if the disabled text
field is filled in. The e-mail gateway should
<emphasis>not</emphasis>
be enabled for secure installations of Bugzilla.</para>
</note>
</para>
</listitem>
<listitem>
<para>
<emphasis>&lt;groupname&gt;</emphasis>:
If you have created some groups, e.g. "securitysensitive", then
checkboxes will appear here to allow you to add users to, or
remove them from, these groups.
</para>
</listitem>
<listitem>
<para>
<emphasis>canconfirm</emphasis>:
This field is only used if you have enabled the "unconfirmed"
status. If you enable this for a user,
that user can then move bugs from "Unconfirmed" to a "Confirmed"
status (e.g.: "New" status).</para>
</listitem>
<listitem>
<para>
<emphasis>creategroups</emphasis>:
This option will allow a user to create and destroy groups in
Bugzilla.</para>
</listitem>
<listitem>
<para>
<emphasis>editbugs</emphasis>:
Unless a user has this bit set, they can only edit those bugs
for which they are the assignee or the reporter. Even if this
option is unchecked, users can still add comments to bugs.
</para>
</listitem>
<listitem>
<para>
<emphasis>editcomponents</emphasis>:
This flag allows a user to create new products and components,
as well as modify and destroy those that have no bugs associated
with them. If a product or component has bugs associated with it,
those bugs must be moved to a different product or component
before Bugzilla will allow them to be destroyed.
</para>
</listitem>
<listitem>
<para>
<emphasis>editkeywords</emphasis>:
If you use Bugzilla's keyword functionality, enabling this
feature allows a user to create and destroy keywords. As always,
the keywords for existing bugs containing the keyword the user
wishes to destroy must be changed before Bugzilla will allow it
to die.</para>
</listitem>
<listitem>
<para>
<emphasis>editusers</emphasis>:
This flag allows a user to do what you're doing right now: edit
other users. This will allow those with the right to do so to
remove administrator privileges from other users or grant them to
themselves. Enable with care.</para>
</listitem>
<listitem>
<para>
<emphasis>tweakparams</emphasis>:
This flag allows a user to change Bugzilla's Params
(using <filename>editparams.cgi</filename>.)</para>
</listitem>
<listitem>
<para>
<emphasis>&lt;productname&gt;</emphasis>:
This allows an administrator to specify the products in which
a user can see bugs. The user must still have the
"editbugs" privilege to edit bugs in these products.</para>
</listitem>
</itemizedlist>
</section>
</section>
</section>
<section id="programadmin">
<title>Product, Component, Milestone, and Version Administration</title>
<section id="products">
<title>Products</title>
<para>
<glossterm linkend="gloss-product" baseform="product">
Products</glossterm>
are the broadest category in Bugzilla, and tend to represent real-world
shipping products. E.g. if your company makes computer games,
you should have one product per game, perhaps a "Common" product for
units of technology used in multiple games, and maybe a few special
products (Website, Administration...)</para>
<para>Many of Bugzilla's settings are configurable on a per-product
basis. The number of "votes" available to users is set per-product,
as is the number of votes
required to move a bug automatically from the UNCONFIRMED status to the
NEW status.</para>
<para>To create a new product:</para>
<orderedlist>
<listitem>
<para>Select "products" from the footer</para>
</listitem>
<listitem>
<para>Select the "Add" link in the bottom right</para>
</listitem>
<listitem>
<para>Enter the name of the product and a description. The
Description field may contain HTML.</para>
</listitem>
</orderedlist>
<para>Don't worry about the "Closed for bug entry", "Maximum Votes
per person", "Maximum votes a person can put on a single bug",
"Number of votes a bug in this Product needs to automatically get out
of the UNCOMFIRMED state", and "Version" options yet. We'll cover
those in a few moments.
</para>
</section>
<section id="components">
<title>Components</title>
<para>Components are subsections of a Product. E.g. the computer game
you are designing may have a "UI"
component, an "API" component, a "Sound System" component, and a
"Plugins" component, each overseen by a different programmer. It
often makes sense to divide Components in Bugzilla according to the
natural divisions of responsibility within your Product or
company.</para>
<para>
Each component has a owner and (if you turned it on in the parameters),
a QA Contact. The owner should be the primary person who fixes bugs in
that component. The QA Contact should be the person who will ensure
these bugs are completely fixed. The Owner, QA Contact, and Reporter
will get email when new bugs are created in this Component and when
these bugs change. Default Owner and Default QA Contact fields only
dictate the
<emphasis>default assignments</emphasis>;
these can be changed on bug submission, or at any later point in
a bug's life.</para>
<para>To create a new Component:</para>
<orderedlist>
<listitem>
<para>Select the "Edit components" link from the "Edit product"
page</para>
</listitem>
<listitem>
<para>Select the "Add" link in the bottom right.</para>
</listitem>
<listitem>
<para>Fill out the "Component" field, a short "Description",
the "Initial Owner" and "Initial QA Contact" (if enabled.)
The Component and Description fields may contain HTML;
the "Initial Owner" field must be a login name
already existing in the database.
</para>
</listitem>
</orderedlist>
</section>
<section id="versions">
<title>Versions</title>
<para>Versions are the revisions of the product, such as "Flinders
3.1", "Flinders 95", and "Flinders 2000". Version is not a multi-select
field; the usual practice is to select the most recent version with
the bug.
</para>
<para>To create and edit Versions:</para>
<orderedlist>
<listitem>
<para>From the "Edit product" screen, select "Edit Versions"</para>
</listitem>
<listitem>
<para>You will notice that the product already has the default
version "undefined". Click the "Add" link in the bottom right.</para>
</listitem>
<listitem>
<para>Enter the name of the Version. This field takes text only.
Then click the "Add" button.</para>
</listitem>
</orderedlist>
</section>
<section id="milestones">
<title>Milestones</title>
<para>Milestones are "targets" that you plan to get a bug fixed by. For
example, you have a bug that you plan to fix for your 3.0 release, it
would be assigned the milestone of 3.0.</para>
<note>
<para>Milestone options will only appear for a Product if you turned
on the "usetargetmilestone" Param in the "Edit Parameters" screen.
</para>
</note>
<para>To create new Milestones, set Default Milestones, and set
Milestone URL:</para>
<orderedlist>
<listitem>
<para>Select "Edit milestones" from the "Edit product" page.</para>
</listitem>
<listitem>
<para>Select "Add" in the bottom right corner.
text</para>
</listitem>
<listitem>
<para>Enter the name of the Milestone in the "Milestone" field. You
can optionally set the "sortkey", which is a positive or negative
number (-255 to 255) that defines where in the list this particular
milestone appears. This is because milestones often do not
occur in alphanumeric order For example, "Future" might be
after "Release 1.2". Select "Add".</para>
</listitem>
<listitem>
<para>From the Edit product screen, you can enter the URL of a
page which gives information about your milestones and what
they mean. </para>
<tip>
<para>If you want your milestone document to be restricted so
that it can only be viewed by people in a particular Bugzilla
group, the best way is to attach the document to a bug in that
group, and make the URL the URL of that attachment.</para>
</tip>
</listitem>
</orderedlist>
</section>
</section>
<section id="voting">
<title>Voting</title>
<para>Voting allows users to be given a pot of votes which they can allocate
to bugs, to indicate that they'd like them fixed.
This allows developers to gauge
user need for a particular enhancement or bugfix. By allowing bugs with
a certain number of votes to automatically move from "UNCONFIRMED" to
"NEW", users of the bug system can help high-priority bugs garner
attention so they don't sit for a long time awaiting triage.</para>
<para>To modify Voting settings:</para>
<orderedlist>
<listitem>
<para>Navigate to the "Edit product" screen for the Product you
wish to modify</para>
</listitem>
<listitem>
<para><emphasis>Maximum Votes per person</emphasis>:
Setting this field to "0" disables voting.</para>
</listitem>
<listitem>
<para><emphasis>Maximum Votes a person can put on a single
bug"</emphasis>:
It should probably be some number lower than the
"Maximum votes per person". Don't set this field to "0" if
"Maximum votes per person" is non-zero; that doesn't make
any sense.</para>
</listitem>
<listitem>
<para><emphasis>Number of votes a bug in this product needs to
automatically get out of the UNCONFIRMED state</emphasis>:
Setting this field to "0" disables the automatic move of
bugs from UNCONFIRMED to NEW.
</para>
</listitem>
<listitem>
<para>Once you have adjusted the values to your preference, click
"Update".</para>
</listitem>
</orderedlist>
</section>
<section id="groups">
<title>Groups and Group Security</title>
<para>Groups allow the administrator
to isolate bugs or products that should only be seen by certain people.
There are two types of group - Generic Groups, and Product-Based Groups.
</para>
<para>
Product-Based Groups are matched with products, and allow you to restrict
access to bugs on a per-product basis. They are enabled using the
usebuggroups Param. Turning on the usebuggroupsentry
Param will mean bugs automatically get added to their product group when
filed.
</para>
<para>
Generic Groups have no special relationship to products;
you create them, and put bugs in them
as required. One example of the use of Generic Groups
is Mozilla's "Security" group,
into which security-sensitive bugs are placed until fixed. Only the
Mozilla Security Team are members of this group.
</para>
<para>To create Generic Groups:</para>
<orderedlist>
<listitem>
<para>Select the "groups"
link in the footer.</para>
</listitem>
<listitem>
<para>Take a moment to understand the instructions on the "Edit
Groups" screen, then select the "Add Group" link.</para>
</listitem>
<listitem>
<para>Fill out the "New Name", "New Description", and
"New User RegExp" fields. "New User RegExp" allows you to automatically
place all users who fulfill the Regular Expression into the new group.
When you have finished, click "Add".</para>
</listitem>
</orderedlist>
<para>To use Product-Based Groups:</para>
<orderedlist>
<listitem>
<para>Turn on "usebuggroups" and "usebuggroupsentry" in the "Edit
Parameters" screen.</para>
<warning>
<para>XXX is this still true?
"usebuggroupsentry" has the capacity to prevent the
administrative user from directly altering bugs because of
conflicting group permissions. If you plan on using
"usebuggroupsentry", you should plan on restricting
administrative account usage to administrative duties only. In
other words, manage bugs with an unpriveleged user account, and
manage users, groups, Products, etc. with the administrative
account.</para>
</warning>
</listitem>
<listitem>
<para>In future, when you create a Product, a matching group will be
automatically created. If you need to add a Product Group to
a Product which was created before you turned on usebuggroups,
then simply create a new group, as outlined above, with the
same name as the Product.</para>
</listitem>
</orderedlist>
<warning>
<para>Bugzilla currently has a limit of 64 groups per installation. If
you have more than about 50 products, you should consider
running multiple Bugzillas. Ask in the newsgroup for other
suggestions for working around this restriction.</para>
</warning>
<para>
Note that group permissions are such that you need to be a member
of <emphasis>all</emphasis> the groups a bug is in, for whatever
reason, to see that bug.
</para>
</section>
<section id="security">
<title>Bugzilla Security</title>
<warning>
<para>Poorly-configured MySQL and Bugzilla installations have
given attackers full access to systems in the past. Please take these
guidelines seriously, even for Bugzilla machines hidden away behind
your firewall. 80% of all computer trespassers are insiders, not
anonymous crackers.</para>
</warning>
<note>
<para>These instructions must, of necessity, be somewhat vague since
Bugzilla runs on so many different platforms. If you have refinements
of these directions, please submit a bug to &bzg-bugs;.
</para>
</note>
<warning>
<para>This is not meant to be a comprehensive list of every possible
security issue regarding the tools mentioned in this section. There is
no subsitute for reading the information written by the authors of any
software running on your system.
</para>
</warning>
<section id="security-networking">
<title>TCP/IP Ports</title>
<!-- TODO: Make this make sense (TCP/IP) -->
<para>TCP/IP defines 65,000 some ports for trafic. Of those, Bugzilla
only needs 1... 2 if you need to use features that require e-mail such
as bug moving or the e-mail interface from contrib. You should audit
your server and make sure that you aren't listening on any ports you
don't need to be. You may also wish to use some kind of firewall
software to be sure that trafic can only be recieved on ports you
specify.
</para>
</section>
<section id="security-mysql">
<title>MySQL</title>
<para>MySQL ships by default with many settings that should be changed.
By defaults it allows anybody to connect from localhost without a
password and have full administrative capabilities. It also defaults to
not have a root password (this is <emphasis>not</emphasis> the same as
the system root). Also, many installations default to running
<application>mysqld</application> as the system root.
</para>
<orderedlist>
<listitem>
<para>Make sure you are running at least version 3.22.32 of MySQL
as earlier versions had notable security holes.
</para>
</listitem>
<listitem>
<para>Consult the documentation that came with your system for
information on making <application>mysqld</application> run as an
unprivleged user.
</para>
</listitem>
<listitem>
<para>You should also be sure to disable the anonymous user account
and set a password for the root user. This is accomplished using the
following commands:
</para>
<programlisting>
<prompt>bash$</prompt> mysql mysql
<prompt>mysql&gt;</prompt> DELETE FROM user WHERE user = '';
<prompt>mysql&gt;</prompt> UPDATE user SET password = password('<replaceable>new_password</replaceable>') WHERE user = 'root';
<prompt>mysql&gt;</prompt> FLUSH PRIVILEGES;
</programlisting>
<para>From this point forward you will need to use
<command>mysql -u root -p</command> and enter
<replaceable>new_password</replaceable> when prompted when using the
mysql client.
</para>
</listitem>
<listitem>
<para>If you run MySQL on the same machine as your httpd server, you
should consider disabling networking from within MySQL by adding
the following to your <filename>/etc/my.conf</filename>:
</para>
<programlisting>
[myslqd]
# Prevent network access to MySQL.
skip-networking
</programlisting>
</listitem>
<listitem>
<para>You may also consider running MySQL, or even all of Bugzilla
in a chroot jail; however, instructions for doing that are beyond
the scope of this document.
</para>
</listitem>
</orderedlist>
</section>
<section id="security-daemon">
<title>Daemon Accounts</title>
<para>Many daemons, such as Apache's httpd and MySQL's mysqld default to
running as either <quote>root</quote> or <quote>nobody</quote>. Running
as <quote>root</quote> introduces obvious security problems, but the
problems introduced by running everything as <quote>nobody</quote> may
not be so obvious. Basically, if you're running every daemon as
<quote>nobody</quote> and one of them gets comprimised, they all get
comprimised. For this reason it is recommended that you create a user
account for each daemon.
</para>
<note>
<para>You will need to set the <varname>webservergroup</varname> to
the group you created for your webserver to run as in
<filename>localconfig</filename>. This will allow
<command>./checksetup.pl</command> to better adjust the file
permissions on your Bugzilla install so as to not require making
anything world-writable.
</para>
</note>
</section>
<section id="security-access">
<title>Web Server Access Controls</title>
<para>There are many files that are placed in the Bugzilla directory
area that should not be accessable from the web. Because of the way
Bugzilla is currently layed out, the list of what should and should
not be accessible is rather complicated. A new installation method
is currently in the works which should solve this by allowing files
that shouldn't be accessible from the web to be placed in directory
outside the webroot. See
<ulink url="http://bugzilla.mozilla.org/show_bug.cgi?id=44659">bug
44659</ulink> for more information.
</para>
<itemizedlist spacing="compact">
<listitem>
<para>In the main Bugzilla directory, you should:</para>
<itemizedlist spacing="compact">
<listitem>
<para>Block:
<simplelist type="inline">
<member><filename>*.pl</filename></member>
<member><filename>*localconfig*</filename></member>
<member><filename>runtests.sh</filename></member>
<member><filename>processmail</filename></member>
<member><filename>syncshadowdb</filename></member>
</simplelist>
</para>
</listitem>
<listitem>
<para>But allow:
<simplelist type="inline">
<member><filename>localconfig.js</filename></member>
<member><filename>localconfig.rdf</filename></member>
</simplelist>
</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>In <filename class="directory">data</filename>:</para>
<itemizedlist spacing="compact">
<listitem>
<para>Block everything</para>
</listitem>
<listitem>
<para>But allow:
<simplelist type="inline">
<member><filename>duplicates.rdf</filename></member>
</simplelist>
</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>In <filename class="directory">data/webdot</filename>:</para>
<itemizedlist spacing="compact">
<listitem>
<para>If you use a remote webdot server:</para>
<itemizedlist spacing="compact">
<listitem>
<para>Block everything</para>
</listitem>
<listitem>
<para>But allow
<simplelist type="inline">
<member><filename>*.dot</filename></member>
</simplelist>
only for the remote webdot server</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>Otherwise, if you use a local GraphViz:</para>
<itemizedlist spacing="compact">
<listitem>
<para>Block everything</para>
</listitem>
<listitem>
<para>But allow:
<simplelist type="inline">
<member><filename>*.png</filename></member>
<member><filename>*.gif</filename></member>
<member><filename>*.jpg</filename></member>
<member><filename>*.map</filename></member>
</simplelist>
</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>And if you don't use any dot:</para>
<itemizedlist spacing="compact">
<listitem>
<para>Block everything</para>
</listitem>
</itemizedlist>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>In <filename class="directory">Bugzilla</filename>:</para>
<itemizedlist spacing="compact">
<listitem>
<para>Block everything</para>
</listitem>
</itemizedlist>
</listitem>
<listitem>
<para>In <filename class="directory">template</filename>:</para>
<itemizedlist spacing="compact">
<listitem>
<para>Block everything</para>
</listitem>
</itemizedlist>
</listitem>
</itemizedlist>
<tip>
<para>Bugzilla ships with the ability to generate
<filename>.htaccess</filename> files instructing Apache which files
should and should not be accessible.
</para>
</tip>
<para>You should test to make sure that the files mentioned above are
not accessible from the Internet, especially your
<filename>localconfig</filename> file which contains your database
password. To test, simply point your web browser at the file; for
example, to test mozilla.org's installation, we'd try to access
<ulink url="http://bugzilla.mozilla.org/localconfig"/>. You should
get a <errorcode>403</errorcode> <errorname>Forbidden</errorname>
error.
</para>
<caution>
<para>Not following the instructions in this section, including
testing, may result in sensitive information being globally
accessible.
</para>
</caution>
</section>
</section>
<section id="cust-templates">
<title>Template Customisation</title>
<para>
One of the large changes for 2.16 was the templatisation of the
entire user-facing UI, using the
<ulink url="http://www.template-toolkit.org">Template Toolkit</ulink>.
Administrators can now configure the look and feel of Bugzilla without
having to edit Perl files or face the nightmare of massive merge
conflicts when they upgrade to a newer version in the future.
</para>
<para>
Templatisation also makes localised versions of Bugzilla possible,
for the first time. In the future, a Bugzilla installation may
have templates installed for multiple localisations, and select
which ones to use based on the user's browser language setting.
</para>
<section>
<title>What to Edit</title>
<para>
There are two different ways of editing of Bugzilla's templates,
and which you use depends mainly on how you upgrade Bugzilla. The
template directory structure is that there's a top level directory,
<filename>template</filename>, which contains a directory for
each installed localisation. The default English templates are
therefore in <filename>en</filename>. Underneath that, there
is the <filename>default</filename> directory and optionally the
<filename>custom</filename> directory. The <filename>default</filename>
directory contains all the templates shipped with Bugzilla, whereas
the <filename>custom</filename> directory does not exist at first and
must be created if you want to use it.
</para>
<para>
The first method of making customisations is to directly edit the
templates in <filename>template/en/default</filename>. This is
probably the best method for small changes if you are going to use
the CVS method of upgrading, because if you then execute a
<command>cvs update</command>, any template fixes will get
automagically merged into your modified versions.
</para>
<para>
If you use this method, your installation will break if CVS conflicts
occur.
</para>
<para>
The other method is to copy the templates into a mirrored directory
structure under <filename>template/en/custom</filename>. The templates
in this directory automatically override those in default.
This is the technique you
need to use if you use the overwriting method of upgrade, because
otherwise your changes will be lost. This method is also better if
you are using the CVS method of upgrading and are going to make major
changes, because it is guaranteed that the contents of this directory
will not be touched during an upgrade, and you can then decide whether
to continue using your own templates, or make the effort to merge your
changes into the new versions by hand.
</para>
<para>
If you use this method, your installation may break if incompatible
changes are made to the template interface. If such changes are made
they will be documented in the release notes, provided you are using a
stable release of Bugzilla. If you use using unstable code, you will
need to deal with this one yourself, although if possible the changes
will be mentioned before they occur in the deprecations section of the
previous stable release's release notes.
</para>
<note>
<para>
Don't directly edit the compiled templates in
<filename class="directory">data/template/*</filename> - your
changes will be lost when Template Toolkit recompiles them.
</para>
</note>
</section>
<section>
<title>How To Edit Templates</title>
<para>
The syntax of the Template Toolkit language is beyond the scope of
this guide. It's reasonably easy to pick up by looking at the current
templates; or, you can read the manual, available on the
<ulink url="http://www.template-toolkit.org">Template Toolkit home
page</ulink>. However, you should particularly remember (for security
reasons) to always HTML filter things which come from the database or
user input, to prevent cross-site scripting attacks.
</para>
<para>
However, one thing you should take particular care about is the need
to properly HTML filter data that has been passed into the template.
This means that if the data can possibly contain special HTML characters
such as &lt;, and the data was not intended to be HTML, they need to be
converted to entity form, ie &amp;lt;. You use the 'html' filter in the
Template Toolkit to do this. If you fail to do this, you may open up
your installation to cross-site scripting attacks.
</para>
<para>
Also note that Bugzilla adds a few filters of its own, that are not
in standard Template Toolkit. In particular, the 'url_quote' filter
can convert characters that are illegal or have special meaning in URLs,
such as &amp;, to the encoded form, ie %26. This actually encodes most
characters (but not the common ones such as letters and numbers and so
on), including the HTML-special characters, so there's never a need to
HTML filter afterwards.
</para>
<para>
Editing templates is a good way of doing a "poor man's custom fields".
For example, if you don't use the Status Whiteboard, but want to have
a free-form text entry box for "Build Identifier", then you can just
edit the templates to change the field labels. It's still be called
status_whiteboard internally, but your users don't need to know that.
</para>
<note>
<para>
If you are making template changes that you intend on submitting back
for inclusion in standard Bugzilla, you should read the relevant
sections of the
<ulink url="http://www.bugzilla.org/developerguide.html">Developers'
Guide</ulink>.
</para>
</note>
</section>
<section>
<title>Template Formats</title>
<para>
Some CGIs have the ability to use more than one template. For
example, buglist.cgi can output bug lists as RDF or two
different forms of HTML (complex and simple). (Try this out
by appending <filename>&amp;format=simple</filename> to a buglist.cgi
URL on your Bugzilla installation.) This
mechanism, called template 'formats', is extensible.
</para>
<para>
To see if a CGI supports multiple output formats, grep the
CGI for "ValidateOutputFormat". If it's not present, adding
multiple format support isn't too hard - see how it's done in
other CGIs.
</para>
<para>
To make a new format template for a CGI which supports this,
open a current template for
that CGI and take note of the INTERFACE comment (if present.) This
comment defines what variables are passed into this template. If
there isn't one, I'm afraid you'll have to read the template and
the code to find out what information you get.
</para>
<para>
Write your template in whatever markup or text style is appropriate.
</para>
<para>
You now need to decide what content type you want your template
served as. Open up the <filename>localconfig</filename> file and find the
<filename>$contenttypes</filename>
variable. If your content type is not there, add it. Remember
the three- or four-letter tag assigned to you content type.
This tag will be part of the template filename.
</para>
<para>
Save the template as <filename>&lt;stubname&gt;-&lt;formatname&gt;.&lt;contenttypetag&gt;.tmpl</filename>.
Try out the template by calling the CGI as
<filename>&lt;cginame&gt;.cgi?format=&lt;formatname&gt;</filename> .
</para>
</section>
<section>
<title>Particular Templates</title>
<para>
There are a few templates you may be particularly interested in
customising for your installation.
</para>
<para>
<command>index.html.tmpl</command>:
This is the Bugzilla front page.
</para>
<para>
<command>global/header.html.tmpl</command>:
This defines the header that goes on all Bugzilla pages.
The header includes the banner, which is what appears to users
and is probably what you want to edit instead. However the
header also includes the HTML HEAD section, so you could for
example add a stylesheet or META tag by editing the header.
</para>
<para>
<command>global/banner.html.tmpl</command>:
This contains the "banner", the part of the header that appears
at the top of all Bugzilla pages. The default banner is reasonably
barren, so you'll probably want to customise this to give your
installation a distinctive look and feel. It is recommended you
preserve the Bugzilla version number in some form so the version
you are running can be determined, and users know what docs to read.
</para>
<para>
<command>global/footer.html.tmpl</command>:
This defines the footer that goes on all Bugzilla pages. Editing
this is another way to quickly get a distinctive look and feel for
your Bugzilla installation.
</para>
<para>
<command>bug/create/user-message.html.tmpl</command>:
This is a message that appears near the top of the bug reporting page.
By modifying this, you can tell your users how they should report
bugs.
</para>
<para>
<command>bug/create/create.html.tmpl</command> and
<command>bug/create/comment.txt.tmpl</command>:
You may wish to get bug submitters to give certain bits of structured
information, each in a separate input widget, for which there is not a
field in the database. The bug entry system has been designed in an
extensible fashion to enable you to define arbitrary fields and widgets,
and have their values appear formatted in the initial
Description, rather than in database fields. An example of this
is the mozilla.org
<ulink url="http://bugzilla.mozilla.org/enter_bug.cgi?format=guided">guided
bug submission form</ulink>.
</para>
<para>
To make this work, create a custom template for
<filename>enter_bug.cgi</filename> (the default template, on which you
could base it, is <filename>create.html.tmpl</filename>),
and either call it <filename>create.html.tmpl</filename> or use a format and
call it <filename>create-&lt;formatname&gt;.html.tmpl</filename>.
Put it in the <filename class="directory">custom/bug/create</filename>
directory. In it, add widgets for each piece of information you'd like
collected - such as a build number, or set of steps to reproduce.
</para>
<para>
Then, create a template like
<filename>custom/bug/create/comment.txt.tmpl</filename>, also named
after your format if you are using one, which
references the form fields you have created. When a bug report is
submitted, the initial comment attached to the bug report will be
formatted according to the layout of this template.
</para>
<para>
For example, if your enter_bug template had a field
<programlisting>&lt;input type="text" name="buildid" size="30"&gt;</programlisting>
and then your comment.txt.tmpl had
<programlisting>BuildID: [% form.buildid %]</programlisting>
then
<programlisting>BuildID: 20020303</programlisting>
would appear in the initial checkin comment.
</para>
</section>
</section>
<section id="upgrading">
<title>Upgrading to New Releases</title>
<para>Upgrading Bugzilla is something we all want to do from time to time,
be it to get new features or pick up the latest security fix. How easy
it is to update depends on a few factors.
</para>
<itemizedlist>
<listitem>
<para>If the new version is a revision or a new point release</para>
</listitem>
<listitem>
<para>How many, if any, local changes have been made</para>
</listitem>
</itemizedlist>
<para>There are also three different methods to upgrade your installation.
</para>
<orderedlist>
<listitem>
<para>Using CVS (<xref linkend="upgrade-cvs"/>)</para>
</listitem>
<listitem>
<para>Downloading a new tarball (<xref linkend="upgrade-tarball"/>)</para>
</listitem>
<listitem>
<para>Applying the relevant patches (<xref linkend="upgrade-patches"/>)</para>
</listitem>
</orderedlist>
<para>Which options are available to you may depend on how large a jump
you are making and/or your network configuration.
</para>
<para>Revisions are normally released to fix security vulnerabilities
and are distinguished by an increase in the third number. For example,
when 2.16.2 was released, it was a revision to 2.16.1.
</para>
<para>Point releases are normally released when the Bugzilla team feels
that there has been a significant amount of progress made between the
last point release and the current time. These are often proceeded by a
stabilization period and release candidates, however the use of
development versions or release candidates is beyond the scope of this
document. Point releases can be distinguished by an increase in the
second number, or minor version. For example, 2.16.2 is a newer point
release than 2.14.5.
</para>
<para>The examples in this section are written as if you were updating
to version 2.16.2. The procedures are the same regardless if you are
updating to a new point release or a new revision. However, the chance
of running into trouble increases when upgrading to a new point release,
escpecially if you've made local changes.
</para>
<para>These examples also assume that your Bugzilla installation is at
<filename>/var/www/html/bugzilla</filename>. If that is not the case,
simply substitute the proper paths where appropriate.
</para>
<example id="upgrade-cvs">
<title>Upgrading using CVS</title>
<para>Every release of Bugzilla, whether it is a revision or a point
release, is tagged in CVS. Also, every tarball we have distributed
since version 2.12 has been primed for using CVS. This does, however,
require that you are able to access cvs-mirror.mozilla.org on port
2401.
<tip>
<para>If you can do this, updating using CVS is probably the most
painless method, especially if you have a lot of local changes.
</para>
</tip>
</para>
<programlisting>
bash$ <command>cd /var/www/html/bugzilla</command>
bash$ <command>cvs login</command>
Logging in to :pserver:anonymous@cvs-mirror.mozilla.org:2401/cvsroot
CVS password: <command>anonymous</command>
bash$ <command>cvs -q update -r BUGZILLA-2_16_2 -dP</command>
P checksetup.pl
P collectstats.pl
P globals.pl
P docs/rel_notes.txt
P template/en/default/list/quips.html.tmpl
</programlisting>
<para>
<caution>
<para>If a line in the output from <command>cvs update</command>
begins with a <computeroutput>C</computeroutput> that represents a
file with local changes that CVS was unable to properly merge. You
need to resolve these conflicts manually before Bugzilla (or at
least the portion using that file) will be usable.
</para>
</caution>
<note>
<para>You also need to run <command>./checksetup.pl</command>
before your Bugzilla upgrade will be complete.
</para>
</note>
</para>
</example>
<example id="upgrade-tarball">
<title>Upgrading using the tarball</title>
<para>If you are unable or unwilling to use CVS, another option that's
always available is to download the latest tarball. This is the most
difficult option to use, especially if you have local changes.
</para>
<programlisting>
bash$ <command>cd /var/www/html</command>
bash$ <command>wget ftp://ftp.mozilla.org/pub/webtools/bugzilla-2.16.2.tar.gz</command>
<emphasis>Output omitted</emphasis>
bash$ <command>tar xzvf bugzilla-2.16.2.tar.gz</command>
bugzilla-2.16.2/
bugzilla-2.16.2/.cvsignore
bugzilla-2.16.2/1x1.gif
<emphasis>Output truncated</emphasis>
bash$ <command>cd bugzilla-2.16.2</command>
bash$ <command>cp ../bugzilla/localconfig* .</command>
bash$ <command>cp -r ../bugzilla/data .</command>
bash$ <command>cd ..</command>
bash$ <command>mv bugzilla bugzilla.old</command>
bash$ <command>mv bugzilla-2.16.2 bugzilla</command>
bash$ <command>cd bugzilla</command>
bash$ <command>./checksetup.pl</command>
<emphasis>Output omitted</emphasis>
</programlisting>
<para>
<warning>
<para>The <command>cp</command> commands both end with periods which
is a very important detail, it tells the shell that the destination
directory is the current working directory. Also, the period at the
beginning of the <command>./checksetup.pl</command> is important and
can not be omitted.
</para>
</warning>
<note>
<para>You will now have to reapply any changes you have made to your
local installation manually.
</para>
</note>
</para>
</example>
<example id="upgrade-patches">
<title>Upgrading using patches</title>
<para>The Bugzilla team will normally make a patch file available for
revisions to go from the most recent revision to the new one. You could
also read the release notes and grab the patches attached to the
mentioned bug, but it is safer to use the released patch file as
sometimes patches get changed before they get checked in (for minor
spelling fixes and the like). It is also theorectically possible to
scour the fixed bug list and pick and choose which patches to apply
from a point release, but this is not recommended either as what you'll
end up with is a hodge podge Bugzilla that isn't really any version.
This would also make it more difficult to upgrade in the future.
</para>
<programlisting>
bash$ <command>cd /var/www/html/bugzilla</command>
bash$ <command>wget ftp://ftp.mozilla.org/pub/webtools/bugzilla-2.16.1-to-2.16.2.diff.gz</command>
<emphasis>Output omitted</emphasis>
bash$ <command>gunzip bugzilla-2.16.1-to-2.16.2.diff.gz</command>
bash$ <command>patch -p1 &lt; bugzilla-2.16.1-to-2.16.2.diff</command>
patching file checksetup.pl
patching file collectstats.pl
patching file globals.pl
</programlisting>
<para>
<caution>
<para>If you do this, beware that this doesn't change the entires in
your <filename id="dir">CVS</filename> directory so it may make
updates using CVS (<xref linkend="upgrade-cvs"/>) more difficult in the
future.
</para>
</caution>
</para>
</example>
</section>
<!-- Integrating Bugzilla with Third-Party Tools -->
&integration;
</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:
-->