run-command: dup_devnull(): guard against syscalls failing
[git/git.git] / Documentation / SubmittingPatches
1 Here are some guidelines for people who want to contribute their code
2 to this software.
3
4 (0) Decide what to base your work on.
5
6 In general, always base your work on the oldest branch that your
7 change is relevant to.
8
9 - A bugfix should be based on 'maint' in general. If the bug is not
10 present in 'maint', base it on 'master'. For a bug that's not yet
11 in 'master', find the topic that introduces the regression, and
12 base your work on the tip of the topic.
13
14 - A new feature should be based on 'master' in general. If the new
15 feature depends on a topic that is in 'pu', but not in 'master',
16 base your work on the tip of that topic.
17
18 - Corrections and enhancements to a topic not yet in 'master' should
19 be based on the tip of that topic. If the topic has not been merged
20 to 'next', it's alright to add a note to squash minor corrections
21 into the series.
22
23 - In the exceptional case that a new feature depends on several topics
24 not in 'master', start working on 'next' or 'pu' privately and send
25 out patches for discussion. Before the final merge, you may have to
26 wait until some of the dependent topics graduate to 'master', and
27 rebase your work.
28
29 - Some parts of the system have dedicated maintainers with their own
30 repositories (see the section "Subsystems" below). Changes to
31 these parts should be based on their trees.
32
33 To find the tip of a topic branch, run "git log --first-parent
34 master..pu" and look for the merge commit. The second parent of this
35 commit is the tip of the topic branch.
36
37 (1) Make separate commits for logically separate changes.
38
39 Unless your patch is really trivial, you should not be sending
40 out a patch that was generated between your working tree and
41 your commit head. Instead, always make a commit with complete
42 commit message and generate a series of patches from your
43 repository. It is a good discipline.
44
45 Give an explanation for the change(s) that is detailed enough so
46 that people can judge if it is good thing to do, without reading
47 the actual patch text to determine how well the code does what
48 the explanation promises to do.
49
50 If your description starts to get too long, that's a sign that you
51 probably need to split up your commit to finer grained pieces.
52 That being said, patches which plainly describe the things that
53 help reviewers check the patch, and future maintainers understand
54 the code, are the most beautiful patches. Descriptions that summarise
55 the point in the subject well, and describe the motivation for the
56 change, the approach taken by the change, and if relevant how this
57 differs substantially from the prior version, are all good things
58 to have.
59
60 Make sure that you have tests for the bug you are fixing.
61
62 When adding a new feature, make sure that you have new tests to show
63 the feature triggers the new behaviour when it should, and to show the
64 feature does not trigger when it shouldn't. Also make sure that the
65 test suite passes after your commit. Do not forget to update the
66 documentation to describe the updated behaviour.
67
68 Oh, another thing. I am picky about whitespaces. Make sure your
69 changes do not trigger errors with the sample pre-commit hook shipped
70 in templates/hooks--pre-commit. To help ensure this does not happen,
71 run git diff --check on your changes before you commit.
72
73
74 (2) Describe your changes well.
75
76 The first line of the commit message should be a short description (50
77 characters is the soft limit, see DISCUSSION in git-commit(1)), and
78 should skip the full stop. It is also conventional in most cases to
79 prefix the first line with "area: " where the area is a filename or
80 identifier for the general area of the code being modified, e.g.
81
82 . archive: ustar header checksum is computed unsigned
83 . git-cherry-pick.txt: clarify the use of revision range notation
84
85 If in doubt which identifier to use, run "git log --no-merges" on the
86 files you are modifying to see the current conventions.
87
88 The body should provide a meaningful commit message, which:
89
90 . explains the problem the change tries to solve, iow, what is wrong
91 with the current code without the change.
92
93 . justifies the way the change solves the problem, iow, why the
94 result with the change is better.
95
96 . alternate solutions considered but discarded, if any.
97
98 Describe your changes in imperative mood, e.g. "make xyzzy do frotz"
99 instead of "[This patch] makes xyzzy do frotz" or "[I] changed xyzzy
100 to do frotz", as if you are giving orders to the codebase to change
101 its behaviour. Try to make sure your explanation can be understood
102 without external resources. Instead of giving a URL to a mailing list
103 archive, summarize the relevant points of the discussion.
104
105
106 (3) Generate your patch using git tools out of your commits.
107
108 git based diff tools generate unidiff which is the preferred format.
109
110 You do not have to be afraid to use -M option to "git diff" or
111 "git format-patch", if your patch involves file renames. The
112 receiving end can handle them just fine.
113
114 Please make sure your patch does not add commented out debugging code,
115 or include any extra files which do not relate to what your patch
116 is trying to achieve. Make sure to review
117 your patch after generating it, to ensure accuracy. Before
118 sending out, please make sure it cleanly applies to the "master"
119 branch head. If you are preparing a work based on "next" branch,
120 that is fine, but please mark it as such.
121
122
123 (4) Sending your patches.
124
125 People on the git mailing list need to be able to read and
126 comment on the changes you are submitting. It is important for
127 a developer to be able to "quote" your changes, using standard
128 e-mail tools, so that they may comment on specific portions of
129 your code. For this reason, all patches should be submitted
130 "inline". If your log message (including your name on the
131 Signed-off-by line) is not writable in ASCII, make sure that
132 you send off a message in the correct encoding.
133
134 WARNING: Be wary of your MUAs word-wrap
135 corrupting your patch. Do not cut-n-paste your patch; you can
136 lose tabs that way if you are not careful.
137
138 It is a common convention to prefix your subject line with
139 [PATCH]. This lets people easily distinguish patches from other
140 e-mail discussions. Use of additional markers after PATCH and
141 the closing bracket to mark the nature of the patch is also
142 encouraged. E.g. [PATCH/RFC] is often used when the patch is
143 not ready to be applied but it is for discussion, [PATCH v2],
144 [PATCH v3] etc. are often seen when you are sending an update to
145 what you have previously sent.
146
147 "git format-patch" command follows the best current practice to
148 format the body of an e-mail message. At the beginning of the
149 patch should come your commit message, ending with the
150 Signed-off-by: lines, and a line that consists of three dashes,
151 followed by the diffstat information and the patch itself. If
152 you are forwarding a patch from somebody else, optionally, at
153 the beginning of the e-mail message just before the commit
154 message starts, you can put a "From: " line to name that person.
155
156 You often want to add additional explanation about the patch,
157 other than the commit message itself. Place such "cover letter"
158 material between the three dash lines and the diffstat. Git-notes
159 can also be inserted using the `--notes` option.
160
161 Do not attach the patch as a MIME attachment, compressed or not.
162 Do not let your e-mail client send quoted-printable. Do not let
163 your e-mail client send format=flowed which would destroy
164 whitespaces in your patches. Many
165 popular e-mail applications will not always transmit a MIME
166 attachment as plain text, making it impossible to comment on
167 your code. A MIME attachment also takes a bit more time to
168 process. This does not decrease the likelihood of your
169 MIME-attached change being accepted, but it makes it more likely
170 that it will be postponed.
171
172 Exception: If your mailer is mangling patches then someone may ask
173 you to re-send them using MIME, that is OK.
174
175 Do not PGP sign your patch, at least for now. Most likely, your
176 maintainer or other people on the list would not have your PGP
177 key and would not bother obtaining it anyway. Your patch is not
178 judged by who you are; a good patch from an unknown origin has a
179 far better chance of being accepted than a patch from a known,
180 respected origin that is done poorly or does incorrect things.
181
182 If you really really really really want to do a PGP signed
183 patch, format it as "multipart/signed", not a text/plain message
184 that starts with '-----BEGIN PGP SIGNED MESSAGE-----'. That is
185 not a text/plain, it's something else.
186
187 Send your patch with "To:" set to the mailing list, with "cc:" listing
188 people who are involved in the area you are touching (the output from
189 "git blame $path" and "git shortlog --no-merges $path" would help to
190 identify them), to solicit comments and reviews.
191
192 After the list reached a consensus that it is a good idea to apply the
193 patch, re-send it with "To:" set to the maintainer [*1*] and "cc:" the
194 list [*2*] for inclusion.
195
196 Do not forget to add trailers such as "Acked-by:", "Reviewed-by:" and
197 "Tested-by:" lines as necessary to credit people who helped your
198 patch.
199
200 [Addresses]
201 *1* The current maintainer: gitster@pobox.com
202 *2* The mailing list: git@vger.kernel.org
203
204
205 (5) Sign your work
206
207 To improve tracking of who did what, we've borrowed the
208 "sign-off" procedure from the Linux kernel project on patches
209 that are being emailed around. Although core GIT is a lot
210 smaller project it is a good discipline to follow it.
211
212 The sign-off is a simple line at the end of the explanation for
213 the patch, which certifies that you wrote it or otherwise have
214 the right to pass it on as a open-source patch. The rules are
215 pretty simple: if you can certify the below:
216
217 Developer's Certificate of Origin 1.1
218
219 By making a contribution to this project, I certify that:
220
221 (a) The contribution was created in whole or in part by me and I
222 have the right to submit it under the open source license
223 indicated in the file; or
224
225 (b) The contribution is based upon previous work that, to the best
226 of my knowledge, is covered under an appropriate open source
227 license and I have the right under that license to submit that
228 work with modifications, whether created in whole or in part
229 by me, under the same open source license (unless I am
230 permitted to submit under a different license), as indicated
231 in the file; or
232
233 (c) The contribution was provided directly to me by some other
234 person who certified (a), (b) or (c) and I have not modified
235 it.
236
237 (d) I understand and agree that this project and the contribution
238 are public and that a record of the contribution (including all
239 personal information I submit with it, including my sign-off) is
240 maintained indefinitely and may be redistributed consistent with
241 this project or the open source license(s) involved.
242
243 then you just add a line saying
244
245 Signed-off-by: Random J Developer <random@developer.example.org>
246
247 This line can be automatically added by git if you run the git-commit
248 command with the -s option.
249
250 Notice that you can place your own Signed-off-by: line when
251 forwarding somebody else's patch with the above rules for
252 D-C-O. Indeed you are encouraged to do so. Do not forget to
253 place an in-body "From: " line at the beginning to properly attribute
254 the change to its true author (see (2) above).
255
256 Also notice that a real name is used in the Signed-off-by: line. Please
257 don't hide your real name.
258
259 If you like, you can put extra tags at the end:
260
261 1. "Reported-by:" is used to credit someone who found the bug that
262 the patch attempts to fix.
263 2. "Acked-by:" says that the person who is more familiar with the area
264 the patch attempts to modify liked the patch.
265 3. "Reviewed-by:", unlike the other tags, can only be offered by the
266 reviewer and means that she is completely satisfied that the patch
267 is ready for application. It is usually offered only after a
268 detailed review.
269 4. "Tested-by:" is used to indicate that the person applied the patch
270 and found it to have the desired effect.
271
272 You can also create your own tag or use one that's in common usage
273 such as "Thanks-to:", "Based-on-patch-by:", or "Mentored-by:".
274
275 ------------------------------------------------
276 Subsystems with dedicated maintainers
277
278 Some parts of the system have dedicated maintainers with their own
279 repositories.
280
281 - git-gui/ comes from git-gui project, maintained by Pat Thoyts:
282
283 git://repo.or.cz/git-gui.git
284
285 - gitk-git/ comes from Paul Mackerras's gitk project:
286
287 git://ozlabs.org/~paulus/gitk
288
289 - po/ comes from the localization coordinator, Jiang Xin:
290
291 https://github.com/git-l10n/git-po/
292
293 Patches to these parts should be based on their trees.
294
295 ------------------------------------------------
296 An ideal patch flow
297
298 Here is an ideal patch flow for this project the current maintainer
299 suggests to the contributors:
300
301 (0) You come up with an itch. You code it up.
302
303 (1) Send it to the list and cc people who may need to know about
304 the change.
305
306 The people who may need to know are the ones whose code you
307 are butchering. These people happen to be the ones who are
308 most likely to be knowledgeable enough to help you, but
309 they have no obligation to help you (i.e. you ask for help,
310 don't demand). "git log -p -- $area_you_are_modifying" would
311 help you find out who they are.
312
313 (2) You get comments and suggestions for improvements. You may
314 even get them in a "on top of your change" patch form.
315
316 (3) Polish, refine, and re-send to the list and the people who
317 spend their time to improve your patch. Go back to step (2).
318
319 (4) The list forms consensus that the last round of your patch is
320 good. Send it to the list and cc the maintainer.
321
322 (5) A topic branch is created with the patch and is merged to 'next',
323 and cooked further and eventually graduates to 'master'.
324
325 In any time between the (2)-(3) cycle, the maintainer may pick it up
326 from the list and queue it to 'pu', in order to make it easier for
327 people play with it without having to pick up and apply the patch to
328 their trees themselves.
329
330 ------------------------------------------------
331 Know the status of your patch after submission
332
333 * You can use Git itself to find out when your patch is merged in
334 master. 'git pull --rebase' will automatically skip already-applied
335 patches, and will let you know. This works only if you rebase on top
336 of the branch in which your patch has been merged (i.e. it will not
337 tell you if your patch is merged in pu if you rebase on top of
338 master).
339
340 * Read the git mailing list, the maintainer regularly posts messages
341 entitled "What's cooking in git.git" and "What's in git.git" giving
342 the status of various proposed changes.
343
344 ------------------------------------------------
345 MUA specific hints
346
347 Some of patches I receive or pick up from the list share common
348 patterns of breakage. Please make sure your MUA is set up
349 properly not to corrupt whitespaces.
350
351 See the DISCUSSION section of git-format-patch(1) for hints on
352 checking your patch by mailing it to yourself and applying with
353 git-am(1).
354
355 While you are at it, check the resulting commit log message from
356 a trial run of applying the patch. If what is in the resulting
357 commit is not exactly what you would want to see, it is very
358 likely that your maintainer would end up hand editing the log
359 message when he applies your patch. Things like "Hi, this is my
360 first patch.\n", if you really want to put in the patch e-mail,
361 should come after the three-dash line that signals the end of the
362 commit message.
363
364
365 Pine
366 ----
367
368 (Johannes Schindelin)
369
370 I don't know how many people still use pine, but for those poor
371 souls it may be good to mention that the quell-flowed-text is
372 needed for recent versions.
373
374 ... the "no-strip-whitespace-before-send" option, too. AFAIK it
375 was introduced in 4.60.
376
377 (Linus Torvalds)
378
379 And 4.58 needs at least this.
380
381 ---
382 diff-tree 8326dd8350be64ac7fc805f6563a1d61ad10d32c (from e886a61f76edf5410573e92e38ce22974f9c40f1)
383 Author: Linus Torvalds <torvalds@g5.osdl.org>
384 Date: Mon Aug 15 17:23:51 2005 -0700
385
386 Fix pine whitespace-corruption bug
387
388 There's no excuse for unconditionally removing whitespace from
389 the pico buffers on close.
390
391 diff --git a/pico/pico.c b/pico/pico.c
392 --- a/pico/pico.c
393 +++ b/pico/pico.c
394 @@ -219,7 +219,9 @@ PICO *pm;
395 switch(pico_all_done){ /* prepare for/handle final events */
396 case COMP_EXIT : /* already confirmed */
397 packheader();
398 +#if 0
399 stripwhitespace();
400 +#endif
401 c |= COMP_EXIT;
402 break;
403
404
405 (Daniel Barkalow)
406
407 > A patch to SubmittingPatches, MUA specific help section for
408 > users of Pine 4.63 would be very much appreciated.
409
410 Ah, it looks like a recent version changed the default behavior to do the
411 right thing, and inverted the sense of the configuration option. (Either
412 that or Gentoo did it.) So you need to set the
413 "no-strip-whitespace-before-send" option, unless the option you have is
414 "strip-whitespace-before-send", in which case you should avoid checking
415 it.
416
417
418 Thunderbird, KMail, GMail
419 -------------------------
420
421 See the MUA-SPECIFIC HINTS section of git-format-patch(1).
422
423 Gnus
424 ----
425
426 '|' in the *Summary* buffer can be used to pipe the current
427 message to an external program, and this is a handy way to drive
428 "git am". However, if the message is MIME encoded, what is
429 piped into the program is the representation you see in your
430 *Article* buffer after unwrapping MIME. This is often not what
431 you would want for two reasons. It tends to screw up non ASCII
432 characters (most notably in people's names), and also
433 whitespaces (fatal in patches). Running 'C-u g' to display the
434 message in raw form before using '|' to run the pipe can work
435 this problem around.