blob: a49113fb936588512ae4b1cd88074a642a87dcc6 [file] [log] [blame]
Paul Jakmafa482832012-03-08 13:51:21 +00001%% -*- mode: text; -*-
2%% $QuaggaId: Format:%an, %ai, %h$ $
3
4\documentclass[oneside]{article}
5\usepackage{parskip}
6\usepackage[bookmarks,colorlinks=true]{hyperref}
7
8\title{Conventions for working on Quagga}
9
10\begin{document}
11\maketitle
12
13This is a living document. Suggestions for updates, via the
14\href{http://lists.quagga.net/mailman/listinfo/quagga-dev}{quagga-dev list},
15are welcome.
16
17\tableofcontents
18
19\section{GUIDELINES FOR HACKING ON QUAGGA}
20\label{sec:guidelines}
21
22
23GNU coding standards apply. Indentation follows the result of
24invoking GNU indent (as of 2.2.8a) with no arguments. Note that this
25uses tabs instead of spaces where possible for leading whitespace, and
26assumes that tabs are every 8 columns. Do not attempt to redefine the
27location of tab stops. Note also that some indentation does not
28follow GNU style. This is a historical accident, and we generally
29only clean up whitespace when code is unmaintainable due to whitespace
30issues, to minimise merging conflicts.
31
32For GNU emacs, use indentation style ``gnu''.
33
34For Vim, use the following lines (note that tabs are at 8, and that
35softtabstop sets the indentation level):
36
37set tabstop=8
38set softtabstop=2
39set shiftwidth=2
40set noexpandtab
41
42Be particularly careful not to break platforms/protocols that you
43cannot test.
44
45New code should have good comments, which explain why the code is correct.
46Changes to existing code should in many cases upgrade the comments when
47necessary for a reviewer to conclude that the change has no unintended
48consequences.
49
50Each file in the Git repository should have a git format-placeholder (like
51an RCS Id keyword), somewhere very near the top, commented out appropriately
52for the file type. The placeholder used for Quagga (replacing <dollar> with
53\$) is:
54
55 \verb|$QuaggaId: <dollar>Format:%an, %ai, %h<dollar> $|
56
57See line 2 of HACKING.tex, the source for this document, for an example.
58
59This placeholder string will be expanded out by the `git archive' commands,
60wihch is used to generate the tar archives for snapshots and releases.
61
62Please document fully the proper use of a new function in the header file
63in which it is declared. And please consult existing headers for
64documentation on how to use existing functions. In particular, please consult
65these header files:
66
67\begin{description}
68 \item{lib/log.h} logging levels and usage guidance
69 \item{[more to be added]}
70\end{description}
71
72If changing an exported interface, please try to deprecate the interface in
73an orderly manner. If at all possible, try to retain the old deprecated
74interface as is, or functionally equivalent. Make a note of when the
75interface was deprecated and guard the deprecated interface definitions in
76the header file, ie:
77
78\begin{verbatim}
79/* Deprecated: 20050406 */
80#if !defined(QUAGGA_NO_DEPRECATED_INTERFACES)
81#warning "Using deprecated <libname> (interface(s)|function(s))"
82...
83#endif /* QUAGGA_NO_DEPRECATED_INTERFACES */
84\end{verbatim}
85
86This is to ensure that the core Quagga sources do not use the deprecated
87interfaces (you should update Quagga sources to use new interfaces, if
88applicable), while allowing external sources to continue to build.
89Deprecated interfaces should be excised in the next unstable cycle.
90
91Note: If you wish, you can test for GCC and use a function
92marked with the 'deprecated' attribute. However, you must provide the
93warning for other compilers.
94
95If changing or removing a command definition, \emph{ensure} that you
96properly deprecate it - use the \_DEPRECATED form of the appropriate DEFUN
97macro. This is \emph{critical}. Even if the command can no longer
98function, you \emph{MUST} still implement it as a do-nothing stub.
99
100Failure to follow this causes grief for systems administrators, as an
101upgrade may cause daemons to fail to start because of unrecognised commands.
102Deprecated commands should be excised in the next unstable cycle. A list of
103deprecated commands should be collated for each release.
104
105See also section~\ref{sec:dll-versioning} below regarding SHARED LIBRARY
106VERSIONING.
107
108
109\section{COMPILE-TIME CONDITIONAL CODE}
110
111Please think very carefully before making code conditional at compile time,
112as it increases maintenance burdens and user confusion. In particular,
113please avoid gratuitious --enable-\ldots switches to the configure script -
114typically code should be good enough to be in Quagga, or it shouldn't be
115there at all.
116
117When code must be compile-time conditional, try have the compiler make it
118conditional rather than the C pre-processor - so that it will still be
119checked by the compiler, even if disabled. I.e. this:
120
121\begin{verbatim}
122 if (SOME_SYMBOL)
123 frobnicate();
124\end{verbatim}
125
126rather than:
127
128\begin{verbatim}
129 #ifdef SOME_SYMBOL
130 frobnicate ();
131 #endif /* SOME_SYMBOL */
132\end{verbatim}
133
134Note that the former approach requires ensuring that SOME\_SYMBOL will be
135defined (watch your AC\_DEFINEs).
136
137
138\section{COMMIT MESSAGES}
139
140The commit message requirements are:
141
142\begin{itemize}
143
144\item The message \emph{MUST} provide a suitable one-line summary followed
145 by a blank line as the very first line of the message, in the form:
146
147 \verb|topic: high-level, one line summary|
148
149 Where topic would tend to be name of a subdirectory, and/or daemon, unless
150 there's a more suitable topic (e.g. 'build'). This topic is used to
151 organise change summaries in release announcements.
152
153\item It should have a suitable "body", which tries to address the
154 following areas, so as to help reviewers and future browsers of the
155 code-base understand why the change is correct (note also the code
156 comment requirements):
157
158 \begin{itemize}
159
160 \item The motivation for the change (does it fix a bug, if so which?
161 add a feature?)
162
163 \item The general approach taken, and trade-offs versus any other
164 approaches.
165
166 \item Any testing undertaken or other information affecting the confidence
167 that can be had in the change.
168
169 \item Information to allow reviewers to be able to tell which specific
170 changes to the code are intended (and hence be able to spot any accidental
171 unintended changes).
172
173 \end{itemize}
174\end{itemize}
175
176The one-line summary must be limited to 54 characters, and all other
177lines to 72 characters.
178
179Commit message bodies in the Quagga project have typically taken the
180following form:
181
182\begin{itemize}
183\item An optional introduction, describing the change generally.
184\item A short description of each specific change made, preferably:
185 \begin{itemize} \item file by file
186 \begin{itemize} \item function by function (use of "ditto", or globs is
187 allowed)
188 \end{itemize}
189 \end{itemize}
190\end{itemize}
191
192Contributors are strongly encouraged to follow this form.
193
194This itemised commit messages allows reviewers to have confidence that the
195author has self-reviewed every line of the patch, as well as providing
196reviewers a clear index of which changes are intended, and descriptions for
197them (C-to-english descriptions are not desireable - some discretion is
198useful). For short patches, a per-function/file break-down may be
199redundant. For longer patches, such a break-down may be essential. A
200contrived example (where the general discussion is obviously somewhat
201redundant, given the one-line summary):
202
203\begin{quote}\begin{verbatim}
204zebra: Enhance frob FSM to detect loss of frob
205
206Add a new DOWN state to the frob state machine to allow the barinator to
207detect loss of frob.
208
209* frob.h: (struct frob) Add DOWN state flag.
210* frob.c: (frob\_change) set/clear DOWN appropriately on state change.
211* bar.c: (barinate) Check frob for DOWN state.
212\end{verbatim}\end{quote}
213
214Please have a look at the git commit logs to get a feel for what the norms
215are.
216
217Note that the commit message format follows git norms, so that ``git
218log --oneline'' will have useful output.
219
220\section{HACKING THE BUILD SYSTEM}
221
222If you change or add to the build system (configure.ac, any Makefile.am,
223etc.), try to check that the following things still work:
224
225\begin{itemize}
226\item make dist
227\item resulting dist tarball builds
228\item out-of-tree builds
229\end{itemize}
230
231The quagga.net site relies on make dist to work to generate snapshots. It
232must work. Common problems are to forget to have some additional file
233included in the dist, or to have a make rule refer to a source file without
234using the srcdir variable.
235
236
237\section{RELEASE PROCEDURE}
238
239\begin{itemize}
240\item Tag the apppropriate commit with a release tag (follow existing
241 conventions).
242
243 [This enables recreating the release, and is just good CM practice.]
244
245\item Create a fresh tar archive of the quagga.net repository, and do a test
246 build:
247
248 \begin{verbatim}
249 git-clone git:///code.quagga.net/quagga.git quagga
250 git-archive --remote=git://code.quagga.net/quagga.git \
251 --prefix=quagga-release/ master | tar -xf -
252 cd quagga-release
253
254 autoreconf -i
255 ./configure
256 make
257 make dist
258 \end{verbatim}
259\end{itemize}
260
261The tarball which `make dist' creates is the tarball to be released! The
262git-archive step ensures you're working with code corresponding to that in
263the official repository, and also carries out keyword expansion. If any
264errors occur, move tags as needed and start over from the fresh checkouts.
265Do not append to tarballs, as this has produced non-standards-conforming
266tarballs in the past.
267
268See also: \url{http://wiki.quagga.net/index.php/Main/Processes}
269
270[TODO: collation of a list of deprecated commands. Possibly can be scripted
271to extract from vtysh/vtysh\_cmd.c]
272
273
274\section{TOOL VERSIONS}
275
276Require versions of support tools are listed in INSTALL.quagga.txt.
277Required versions should only be done with due deliberation, as it can
278cause environments to no longer be able to compile quagga.
279
280
281\section{SHARED LIBRARY VERSIONING}
282\label{sec:dll-versioning}
283
284[this section is at the moment just gdt's opinion]
285
286Quagga builds several shared libaries (lib/libzebra, ospfd/libospf,
287ospfclient/libsopfapiclient). These may be used by external programs,
288e.g. a new routing protocol that works with the zebra daemon, or
289ospfapi clients. The libtool info pages (node Versioning) explain
290when major and minor version numbers should be changed. These values
291are set in Makefile.am near the definition of the library. If you
292make a change that requires changing the shared library version,
293please update Makefile.am.
294
295libospf exports far more than it should, and is needed by ospfapi
296clients. Only bump libospf for changes to functions for which it is
297reasonable for a user of ospfapi to call, and please err on the side
298of not bumping.
299
300There is no support intended for installing part of zebra. The core
301library libzebra and the included daemons should always be built and
302installed together.
303
304
305\section{GIT COMMIT SUBMISSION}
306\label{sec:git-submission}
307
308The preferred method for submitting changes is to provide git commits via a
309publically-accessible git repository, which the maintainers can easily pull.
310
311The commits should be in a branch based off the Quagga.net master - a
312"feature branch". Ideally there should be no commits to this branch other
313than those in master, and those intended to be submitted. However, merge
314commits to this branch from the Quagga master are permitted, though strongly
315discouraged - use another (potentially local and throw-away) branch to test
316merge with the latest Quagga master.
317
318Recommended practice is to keep different logical sets of changes on
319separate branches - "topic" or "feature" branches. This allows you to still
320merge them together to one branch (potentially local and/or "throw-away")
321for testing or use, while retaining smaller, independent branches that are
322easier to merge.
323
324All content guidelines in section \ref{sec:patch-submission}, PATCH
325SUBMISSION apply.
326
327
328\section{PATCH SUBMISSION}
329\label{sec:patch-submission}
330
331\begin{itemize}
332
333\item For complex changes, contributors are strongly encouraged to first
334 start a design discussion on the quagga-dev list \emph{before}
335 starting any coding.
336
337\item Send a clean diff against the 'master' branch of the quagga.git
338 repository, in unified diff format, preferably with the '-p' argument to
339 show C function affected by any chunk, and with the -w and -b arguments to
340 minimise changes. E.g:
341
342 git diff -up mybranch..remotes/quagga.net/master
343
344 It is preferable to use git format-patch, and even more preferred to
345 publish a git repository (see GIT COMMIT SUBMISSION, section
346 \ref{sec:git-submission}).
347
348 If not using git format-patch, Include the commit message in the email.
349
350\item After a commit, code should have comments explaining to the reviewer
351 why it is correct, without reference to history. The commit message
352 should explain why the change is correct.
353
354\item Include NEWS entries as appropriate.
355
356\item Include only one semantic change or group of changes per patch.
357
358\item Do not make gratuitous changes to whitespace. See the w and b arguments
359 to diff.
360
361\item Changes should be arranged so that the least contraversial and most
362 trivial are first, and the most complex or more contraversial are
363 last. This will maximise how many the Quagga maintainers can merge,
364 even if some other commits need further work.
365
366\item Providing a unit-test is strongly encouraged. Doing so will make it
367 much easier for maintainers to have confidence that they will be able
368 to support your change.
369
370\item New code should be arranged so that it easy to verify and test. E.g.
371 stateful logic should be separated out from functional logic as much as
372 possible: wherever possible, move complex logic out to smaller helper
373 functions which access no state other than their arguments.
374
375\item State on which platforms and with what daemons the patch has been
376 tested. Understand that if the set of testing locations is small,
377 and the patch might have unforeseen or hard to fix consequences that
378 there may be a call for testers on quagga-dev, and that the patch
379 may be blocked until test results appear.
380
381 If there are no users for a platform on quagga-dev who are able and
382 willing to verify -current occasionally, that platform may be
383 dropped from the "should be checked" list.
384
385\end{itemize}
386
387\section{PATCH APPLICATION}
388
389\begin{itemize}
390
391\item Only apply patches that meet the submission guidelines.
392
393\item If the patch might break something, issue a call for testing on the
394 mailinglist.
395
396\item Give an appropriate commit message (see above), and use the --author
397 argument to git-commit, if required, to ensure proper attribution (you
398 should still be listed as committer)
399
400\item Immediately after commiting, double-check (with git-log and/or gitk).
401 If there's a small mistake you can easily fix it with `git commit
402 --amend ..'
403
404\item When merging a branch, always use an explicit merge commit. Giving
405 --no-ff ensures a merge commit is created which documents ``this human
406 decided to merge this branch at this time''.
407\end{itemize}
408
409\section{STABLE PLATFORMS AND DAEMONS}
410
411The list of platforms that should be tested follow. This is a list
412derived from what quagga is thought to run on and for which
413maintainers can test or there are people on quagga-dev who are able
414and willing to verify that -current does or does not work correctly.
415
416\begin{itemize}
417 \item BSD (Free, Net or Open, any platform)
418 \item GNU/Linux (any distribution, i386)
419 \item Solaris (strict alignment, any platform)
420 \item future: NetBSD/sparc64
421\end{itemize}
422
423The list of daemons that are thought to be stable and that should be
424tested are:
425
426\begin{itemize}
427 \item zebra
428 \item bgpd
429 \item ripd
430 \item ospfd
431 \item ripngd
432\end{itemize}
433Daemons which are in a testing phase are
434
435\begin{itemize}
436 \item ospf6d
437 \item isisd
438 \item watchquagga
439\end{itemize}
440
441\section{IMPORT OR UPDATE VENDOR SPECIFIC ROUTING PROTOCOLS}
442
443The source code of Quagga is based on two vendors:
444
445 \verb|zebra_org| (\url{http://www.zebra.org/})
446 \verb|isisd_sf| (\url{http://isisd.sf.net/})
447
448To import code from further sources, e.g. for archival purposes without
449necessarily having to review and/or fix some changeset, create a branch from
450`master':
451
452\begin{verbatim}
453 git checkout -b archive/foo master
454 <apply changes>
455 git commit -a "Joe Bar <joe@example.com>"
456 git push quagga archive/foo
457\end{verbatim}
458
459presuming `quagga' corresponds to a file in your .git/remotes with
460configuration for the appropriate Quagga.net repository.
461
462\end{document}