evolution/HACKING

1 Patch guidelines

This section lists some guidelines for writing a good patch which is
more likely to be accepted.

Any new features or large scale work should first be discussed on the
evolution-hackers list first.  This will ensure the idea fits in the
direction we wish to take Evolution, and also that the effort is not
duplicated.  See section 3 for details on the mailing lists.

1.1 Patch basics

o The patch should apply cleanly at the time it is made.

o It must compile once applied.

o It must not generate any more compile time warnings than were
  already there.  This may be platform dependent so simply do your
  best.

o It must conform to C89/C90 (ANSI/ISO C), and build with gcc using
  the default compile flags.

  The primary trap is that in C99 you may define variables anywhere in
  the code, in C89 they must be declared in a declaration block which
  follows any block start '{'.

  If you wish to ensure the code is C89, try the following.

  From the gcc manual page:
                                                           "To select
  this standard in GCC, use one of the options `-ansi', `-std=c89' or
   `-std=iso9899:1990'; to obtain all the diagnostics required by the
  standard, you should also specify `-pedantic'" ...

  You may actually have to use '-std=gnu89' if libraries have taken
  advantage of gcc extensions and where not compiled similarly, as the
  above options will disable all gnu extensions.

  [FIXME: Add the same option for Forte here]

o It should not add any extra debug printing by default, unless the
  patch is specifically to add extra debug printing.

o It should not use any gcc extensions, except where they are properly
  checked for and not used with other compilers.  glib provides some
  of these features as portable macros and should be used when they
  cover the required functionality.

1.1 GUI changes

If the change requires non-trivial user interface changes, then they
will have to be discussed and approved on the evolution-hackers list
first.  This is highly recommended before embarking on any UI work, or
large scale work in general.  The Gnome HIG document is the place to
start on any UI changes or additions.

1.2 Translated string changes

Any changes to translated strings in a stable release must be
discussed on the hackers list (see section 3), and/or as part of the
patch submission.  There must be very good reasons for changing the
strings in this case.

1.3 Coding style

Generally the coding style employed matches the "Linux Kernel" style,
that is, basically K&R style indenting with 8 space tabs.  Tabs should
be used rather than space characters.  Reformatting of otherwise
unchanged code is not acceptable.  Editors should have any automatic
reformatting features disabled.

K&R style indenting puts braces on the same line.  The opening
parenthesis of a function call or conditional statement should be on
the same line as the function.  "else" "} else" and "} else {" must
always occur on lines by themselves.

A single blank line should follow {} blocks (if not immediately
followed by the close of another block), and conditional statements,
and be used to separate logical groups of statements in the same
block.

A single blank line only should separate functions, and other
structures at the top level of the file (i.e. outside functions).  The
same rule applies to variable declarations at the start of a block.

An example of the most-developer-preferred formatting:

TheType
the_function (int frank)
{
        int a = 1;

        if (a == frank) {
                a = foo (a);
        } else {
                do {
                        a = bob (frank) + a;
        	} until (a == frank);

                frank = a;
        }

        return (TheType) a;
}

Where there are slight stylistic differences, the style in the
surrounding code should be followed.

1.3.1 Object casts

You can either use C style casts, or Gtk style casts.  Note that Gtk
style casts can add significant execution overhead, which is not
adding any extra checking.  e.g. if arguments have already been
type-checked by preconditions.  Putting a space between a cast and a
variable is optional, but preferred by most of the developers.

1.3.2 Preconditions

External api entry points should have preconditions (g_return_if_fail,
etc), although their use varies from case to case.  Internal entry
points and/or when you are guaranteed the type has already been
checked, are unecessary.  Object initialisation and other virtual
method invocations are considered internal entry points.

1.3.3 Line lengths

Do not expend effort and resort to unreadable formatting merely to fit
any long lines into 80 column widths.  We use 8 space tabs, and
because of the lack of namespacing other than extending the function
name, many of the function and type names are too long for this to be
practical.  We now all uses high resolution displays, and not
circa-80's VT100 terminals!

On the other hand, lines should generally not exceed 100 characters,
and absolutely not exceed 160 characters.  If your tab nesting is too
deep you probably have a poor design that needs rethinking.

1.4 Design

This is a tricky issue to document, but the design of new code should
`fit' with the existing design of the relevent module.  It should at
the very least, be no worse.

Code should not cross existing abstraction boundaries or attempt
to remove or work around them, if required the existing design may
need adjustment.

Type and method names should follow the existing practice in the
surrounding code.  Method arguments should follow the same order as
related methods, and should use the same names for matching
parameters.

Per file, static class globals are ok, true globals may be ok, but
should be used sparingly.  Use 'i' for a loop variable, if that's all
it is, don't use 'the_current_index'.  etc.

If in doubt, ask on the lists.

2. Patch submission guidelines

This section outlines procedures that should be followed when
submitting patches to evolution, via the evolution-patches mailing
list.

You must subcribe to the list at
`http://lists.ximian.com/mailman/listinfo/evolution-patches' before you
can submit patches to it.

Also note that if you attach a patch to a bug report, it should always
be sent to the list for attention.

Any non-trival patches (patches of more than 1 or 2 changed lines in
more than 5 isolated locations) also require copyright assignment.
See http://developer.ximian.com/projects/evolution/copyright.html for
details.

If you follow the guidelines listed here, you should generally expect
a response within 2 working days.  If you re-send the same patch
repeatedly, you will more likely receive less attention.  Do not
re-send the same patch repeatedly.

2.1 Subject Lines

If the patch addresses a specific bug in bugzilla.ximian.com, then the
bug number must be included in the subject line, preferably near the
beginning of the subject line.  A concise summary of the bug(s) being
addressed, should be the remainder of the subject.

It is unnecessary to add "[PATCH]", "patch" or similar to the subject
line, unless it is being cross-posted to other non-patch lists.

It is absolutely unnecessary to add "please consider", "please review",
or "seeking review", or similar, to the subject line.  Please do not do
this.

Where the patch does not address a specific bug number, then the subject
line should simply be a concise summary of the problem/feature it
addresses.

In all cases the subject line should include the module(s) to which the
patch applies, and would generally match the component on the bug or
the top-level module directory (e.g. camel, mail, addressbook, use 'all'
for more than 3 or 4 modules).

2.2 Message Body

Patches should be attached as attachments, preferably as a single
diff, when possible, and the changes are related.  The diff must be in
unified diff format, "-up" is a suitable argument to give to "cvs
diff" (-p may be dropped if not supported by your diff).  If you have
added files, then -N should also be used, but if you are using cvs,
"cvs add" is needed, and requires write access to the repository.

If the patch does not address a specific bug, then the patch email
should describe which feature or problem it addresses.  If it does
address a specific bug, then further explanation beyond the bug
commentary is optional, although often convenient.

It would also be helpful to summarise the module to which it applies
in the message body.

In all cases you should include which branch, or branches, the patch
is intended to apply to.  If this is not given it will be assumed to
be the trunk (HEAD), and such patches will and must not be applied to
any stable branch without further approval.

2.3 ChangeLogs

All patches must include appropriate ChangeLog diff's, to the
appropriate ChangeLog(s) for the given change (emacs will automatically
find the correct one, and format the entry appropriately).  All but
the most trivial of patches will not be considered or discussed
without this.  It is ok to contain extra ChangeLog entries for other
pending patches, but they should not be excessively long - it isn't
that hard to isolate patch diffs.  If the patch addresses a bug in
bugzilla.ximian.com, then the ChangeLog entry must include some
reference to that bug number (either the number, or #number, or 'bug
xxx').  If it addresses a bug in another bug system, it must also
indicate which bug system ('gnome bugzilla' 'red-hat bugzilla', etc).

2.4 Stable branches

Generally, any patch to the stable branch from non-core developers
must address a specific bug in bugzilla.ximian.com.  The patch should
also be attached to the bug in question, with the keyword 'patch' set
on the bug report.  The patch email must identify which stable branch
and version it is to apply to.

3 Mailing lists

3.1 Evolution Hackers

If you wish to discuss patches before they are submitted, or ideas
before you start to work on them, do it on the evolution-hackers list,
which may be subscribed and viewed at
`http://lists.ximian.com/mailman/listinfo/evolution-hackers'.

This is a low-volume list (5-10 posts per day on average).

Some patches may be discussed here to get a wider audience, although
once a patch has been made it should generally be discussed on
evolution-patches.

Feature requests, bug reports, and other user related discussions,
without the intention to write code to address them, will be ignored.

3.2 Evolution Patches

The patch submission list evolution-patches may be subscribed and
viewed at
`http://lists.ximian.com/mailman/listinfo/evolution-patches'.  Once a
patch has been written, it may be submitted here for discussion, as
well as final approval.

Any non-patch related postings to this list will be ignored.