SubmittingPatches: itemize and reflect upon well written changes
[git/git.git] / Documentation / SubmittingPatches
1 Checklist (and a short version for the impatient):
2
3 Commits:
4
5 - make commits of logical units
6 - check for unnecessary whitespace with "git diff --check"
7 before committing
8 - do not check in commented out code or unneeded files
9 - the first line of the commit message should be a short
10 description and should skip the full stop
11 - the body should provide a meaningful commit message, which:
12 - uses the imperative, present tense: "change",
13 not "changed" or "changes".
14 - includes motivation for the change, and contrasts
15 its implementation with previous behaviour
16 - if you want your work included in git.git, add a
17 "Signed-off-by: Your Name <you@example.com>" line to the
18 commit message (or just use the option "-s" when
19 committing) to confirm that you agree to the Developer's
20 Certificate of Origin
21 - make sure that you have tests for the bug you are fixing
22 - make sure that the test suite passes after your commit
23
24 Patch:
25
26 - use "git format-patch -M" to create the patch
27 - do not PGP sign your patch
28 - do not attach your patch, but read in the mail
29 body, unless you cannot teach your mailer to
30 leave the formatting of the patch alone.
31 - be careful doing cut & paste into your mailer, not to
32 corrupt whitespaces.
33 - provide additional information (which is unsuitable for
34 the commit message) between the "---" and the diffstat
35 - if you change, add, or remove a command line option or
36 make some other user interface change, the associated
37 documentation should be updated as well.
38 - if your name is not writable in ASCII, make sure that
39 you send off a message in the correct encoding.
40 - send the patch to the list (git@vger.kernel.org) and the
41 maintainer (gitster@pobox.com) if (and only if) the patch
42 is ready for inclusion. If you use git-send-email(1),
43 please test it first by sending email to yourself.
44
45 Long version:
46
47 I started reading over the SubmittingPatches document for Linux
48 kernel, primarily because I wanted to have a document similar to
49 it for the core GIT to make sure people understand what they are
50 doing when they write "Signed-off-by" line.
51
52 But the patch submission requirements are a lot more relaxed
53 here on the technical/contents front, because the core GIT is
54 thousand times smaller ;-). So here is only the relevant bits.
55
56
57 (1) Make separate commits for logically separate changes.
58
59 Unless your patch is really trivial, you should not be sending
60 out a patch that was generated between your working tree and
61 your commit head. Instead, always make a commit with complete
62 commit message and generate a series of patches from your
63 repository. It is a good discipline.
64
65 Describe the technical detail of the change(s).
66
67 If your description starts to get too long, that's a sign that you
68 probably need to split up your commit to finer grained pieces.
69 That being said, patches which plainly describe the things that
70 help reviewers check the patch, and future maintainers understand
71 the code, are the most beautiful patches. Descriptions that summarise
72 the point in the subject well, and describe the motivation for the
73 change, the approach taken by the change, and if relevant how this
74 differs substantially from the prior version, can be found on Usenet
75 archives back into the late 80's. Consider it like good Netiquette,
76 but for code.
77
78 Oh, another thing. I am picky about whitespaces. Make sure your
79 changes do not trigger errors with the sample pre-commit hook shipped
80 in templates/hooks--pre-commit. To help ensure this does not happen,
81 run git diff --check on your changes before you commit.
82
83
84 (1a) Try to be nice to older C compilers
85
86 We try to support a wide range of C compilers to compile
87 git with. That means that you should not use C99 initializers, even
88 if a lot of compilers grok it.
89
90 Also, variables have to be declared at the beginning of the block
91 (you can check this with gcc, using the -Wdeclaration-after-statement
92 option).
93
94 Another thing: NULL pointers shall be written as NULL, not as 0.
95
96
97 (2) Generate your patch using git tools out of your commits.
98
99 git based diff tools (git, Cogito, and StGIT included) generate
100 unidiff which is the preferred format.
101
102 You do not have to be afraid to use -M option to "git diff" or
103 "git format-patch", if your patch involves file renames. The
104 receiving end can handle them just fine.
105
106 Please make sure your patch does not include any extra files
107 which do not belong in a patch submission. Make sure to review
108 your patch after generating it, to ensure accuracy. Before
109 sending out, please make sure it cleanly applies to the "master"
110 branch head. If you are preparing a work based on "next" branch,
111 that is fine, but please mark it as such.
112
113
114 (3) Sending your patches.
115
116 People on the git mailing list need to be able to read and
117 comment on the changes you are submitting. It is important for
118 a developer to be able to "quote" your changes, using standard
119 e-mail tools, so that they may comment on specific portions of
120 your code. For this reason, all patches should be submitted
121 "inline". WARNING: Be wary of your MUAs word-wrap
122 corrupting your patch. Do not cut-n-paste your patch; you can
123 lose tabs that way if you are not careful.
124
125 It is a common convention to prefix your subject line with
126 [PATCH]. This lets people easily distinguish patches from other
127 e-mail discussions. Use of additional markers after PATCH and
128 the closing bracket to mark the nature of the patch is also
129 encouraged. E.g. [PATCH/RFC] is often used when the patch is
130 not ready to be applied but it is for discussion, [PATCH v2],
131 [PATCH v3] etc. are often seen when you are sending an update to
132 what you have previously sent.
133
134 "git format-patch" command follows the best current practice to
135 format the body of an e-mail message. At the beginning of the
136 patch should come your commit message, ending with the
137 Signed-off-by: lines, and a line that consists of three dashes,
138 followed by the diffstat information and the patch itself. If
139 you are forwarding a patch from somebody else, optionally, at
140 the beginning of the e-mail message just before the commit
141 message starts, you can put a "From: " line to name that person.
142
143 You often want to add additional explanation about the patch,
144 other than the commit message itself. Place such "cover letter"
145 material between the three dash lines and the diffstat.
146
147 Do not attach the patch as a MIME attachment, compressed or not.
148 Do not let your e-mail client send quoted-printable. Do not let
149 your e-mail client send format=flowed which would destroy
150 whitespaces in your patches. Many
151 popular e-mail applications will not always transmit a MIME
152 attachment as plain text, making it impossible to comment on
153 your code. A MIME attachment also takes a bit more time to
154 process. This does not decrease the likelihood of your
155 MIME-attached change being accepted, but it makes it more likely
156 that it will be postponed.
157
158 Exception: If your mailer is mangling patches then someone may ask
159 you to re-send them using MIME, that is OK.
160
161 Do not PGP sign your patch, at least for now. Most likely, your
162 maintainer or other people on the list would not have your PGP
163 key and would not bother obtaining it anyway. Your patch is not
164 judged by who you are; a good patch from an unknown origin has a
165 far better chance of being accepted than a patch from a known,
166 respected origin that is done poorly or does incorrect things.
167
168 If you really really really really want to do a PGP signed
169 patch, format it as "multipart/signed", not a text/plain message
170 that starts with '-----BEGIN PGP SIGNED MESSAGE-----'. That is
171 not a text/plain, it's something else.
172
173 Note that your maintainer does not necessarily read everything
174 on the git mailing list. If your patch is for discussion first,
175 send it "To:" the mailing list, and optionally "cc:" him. If it
176 is trivially correct or after the list reached a consensus, send
177 it "To:" the maintainer and optionally "cc:" the list for
178 inclusion.
179
180 Also note that your maintainer does not actively involve himself in
181 maintaining what are in contrib/ hierarchy. When you send fixes and
182 enhancements to them, do not forget to "cc: " the person who primarily
183 worked on that hierarchy in contrib/.
184
185
186 (4) Sign your work
187
188 To improve tracking of who did what, we've borrowed the
189 "sign-off" procedure from the Linux kernel project on patches
190 that are being emailed around. Although core GIT is a lot
191 smaller project it is a good discipline to follow it.
192
193 The sign-off is a simple line at the end of the explanation for
194 the patch, which certifies that you wrote it or otherwise have
195 the right to pass it on as a open-source patch. The rules are
196 pretty simple: if you can certify the below:
197
198 Developer's Certificate of Origin 1.1
199
200 By making a contribution to this project, I certify that:
201
202 (a) The contribution was created in whole or in part by me and I
203 have the right to submit it under the open source license
204 indicated in the file; or
205
206 (b) The contribution is based upon previous work that, to the best
207 of my knowledge, is covered under an appropriate open source
208 license and I have the right under that license to submit that
209 work with modifications, whether created in whole or in part
210 by me, under the same open source license (unless I am
211 permitted to submit under a different license), as indicated
212 in the file; or
213
214 (c) The contribution was provided directly to me by some other
215 person who certified (a), (b) or (c) and I have not modified
216 it.
217
218 (d) I understand and agree that this project and the contribution
219 are public and that a record of the contribution (including all
220 personal information I submit with it, including my sign-off) is
221 maintained indefinitely and may be redistributed consistent with
222 this project or the open source license(s) involved.
223
224 then you just add a line saying
225
226 Signed-off-by: Random J Developer <random@developer.example.org>
227
228 This line can be automatically added by git if you run the git-commit
229 command with the -s option.
230
231 Notice that you can place your own Signed-off-by: line when
232 forwarding somebody else's patch with the above rules for
233 D-C-O. Indeed you are encouraged to do so. Do not forget to
234 place an in-body "From: " line at the beginning to properly attribute
235 the change to its true author (see (2) above).
236
237 Also notice that a real name is used in the Signed-off-by: line. Please
238 don't hide your real name.
239
240 Some people also put extra tags at the end.
241
242 "Acked-by:" says that the patch was reviewed by the person who
243 is more familiar with the issues and the area the patch attempts
244 to modify. "Tested-by:" says the patch was tested by the person
245 and found to have the desired effect.
246
247 ------------------------------------------------
248 An ideal patch flow
249
250 Here is an ideal patch flow for this project the current maintainer
251 suggests to the contributors:
252
253 (0) You come up with an itch. You code it up.
254
255 (1) Send it to the list and cc people who may need to know about
256 the change.
257
258 The people who may need to know are the ones whose code you
259 are butchering. These people happen to be the ones who are
260 most likely to be knowledgeable enough to help you, but
261 they have no obligation to help you (i.e. you ask for help,
262 don't demand). "git log -p -- $area_you_are_modifying" would
263 help you find out who they are.
264
265 (2) You get comments and suggestions for improvements. You may
266 even get them in a "on top of your change" patch form.
267
268 (3) Polish, refine, and re-send to the list and the people who
269 spend their time to improve your patch. Go back to step (2).
270
271 (4) The list forms consensus that the last round of your patch is
272 good. Send it to the list and cc the maintainer.
273
274 (5) A topic branch is created with the patch and is merged to 'next',
275 and cooked further and eventually graduates to 'master'.
276
277 In any time between the (2)-(3) cycle, the maintainer may pick it up
278 from the list and queue it to 'pu', in order to make it easier for
279 people play with it without having to pick up and apply the patch to
280 their trees themselves.
281
282 ------------------------------------------------
283 MUA specific hints
284
285 Some of patches I receive or pick up from the list share common
286 patterns of breakage. Please make sure your MUA is set up
287 properly not to corrupt whitespaces. Here are two common ones
288 I have seen:
289
290 * Empty context lines that do not have _any_ whitespace.
291
292 * Non empty context lines that have one extra whitespace at the
293 beginning.
294
295 One test you could do yourself if your MUA is set up correctly is:
296
297 * Send the patch to yourself, exactly the way you would, except
298 To: and Cc: lines, which would not contain the list and
299 maintainer address.
300
301 * Save that patch to a file in UNIX mailbox format. Call it say
302 a.patch.
303
304 * Try to apply to the tip of the "master" branch from the
305 git.git public repository:
306
307 $ git fetch http://kernel.org/pub/scm/git/git.git master:test-apply
308 $ git checkout test-apply
309 $ git reset --hard
310 $ git am a.patch
311
312 If it does not apply correctly, there can be various reasons.
313
314 * Your patch itself does not apply cleanly. That is _bad_ but
315 does not have much to do with your MUA. Please rebase the
316 patch appropriately.
317
318 * Your MUA corrupted your patch; "am" would complain that
319 the patch does not apply. Look at .git/rebase-apply/ subdirectory and
320 see what 'patch' file contains and check for the common
321 corruption patterns mentioned above.
322
323 * While you are at it, check what are in 'info' and
324 'final-commit' files as well. If what is in 'final-commit' is
325 not exactly what you would want to see in the commit log
326 message, it is very likely that your maintainer would end up
327 hand editing the log message when he applies your patch.
328 Things like "Hi, this is my first patch.\n", if you really
329 want to put in the patch e-mail, should come after the
330 three-dash line that signals the end of the commit message.
331
332
333 Pine
334 ----
335
336 (Johannes Schindelin)
337
338 I don't know how many people still use pine, but for those poor
339 souls it may be good to mention that the quell-flowed-text is
340 needed for recent versions.
341
342 ... the "no-strip-whitespace-before-send" option, too. AFAIK it
343 was introduced in 4.60.
344
345 (Linus Torvalds)
346
347 And 4.58 needs at least this.
348
349 ---
350 diff-tree 8326dd8350be64ac7fc805f6563a1d61ad10d32c (from e886a61f76edf5410573e92e38ce22974f9c40f1)
351 Author: Linus Torvalds <torvalds@g5.osdl.org>
352 Date: Mon Aug 15 17:23:51 2005 -0700
353
354 Fix pine whitespace-corruption bug
355
356 There's no excuse for unconditionally removing whitespace from
357 the pico buffers on close.
358
359 diff --git a/pico/pico.c b/pico/pico.c
360 --- a/pico/pico.c
361 +++ b/pico/pico.c
362 @@ -219,7 +219,9 @@ PICO *pm;
363 switch(pico_all_done){ /* prepare for/handle final events */
364 case COMP_EXIT : /* already confirmed */
365 packheader();
366 +#if 0
367 stripwhitespace();
368 +#endif
369 c |= COMP_EXIT;
370 break;
371
372
373 (Daniel Barkalow)
374
375 > A patch to SubmittingPatches, MUA specific help section for
376 > users of Pine 4.63 would be very much appreciated.
377
378 Ah, it looks like a recent version changed the default behavior to do the
379 right thing, and inverted the sense of the configuration option. (Either
380 that or Gentoo did it.) So you need to set the
381 "no-strip-whitespace-before-send" option, unless the option you have is
382 "strip-whitespace-before-send", in which case you should avoid checking
383 it.
384
385
386 Thunderbird
387 -----------
388
389 (A Large Angry SCM)
390
391 By default, Thunderbird will both wrap emails as well as flag them as
392 being 'format=flowed', both of which will make the resulting email unusable
393 by git.
394
395 Here are some hints on how to successfully submit patches inline using
396 Thunderbird.
397
398 There are two different approaches. One approach is to configure
399 Thunderbird to not mangle patches. The second approach is to use
400 an external editor to keep Thunderbird from mangling the patches.
401
402 Approach #1 (configuration):
403
404 This recipe is current as of Thunderbird 2.0.0.19. Three steps:
405 1. Configure your mail server composition as plain text
406 Edit...Account Settings...Composition & Addressing,
407 uncheck 'Compose Messages in HTML'.
408 2. Configure your general composition window to not wrap
409 Edit..Preferences..Composition, wrap plain text messages at 0
410 3. Disable the use of format=flowed
411 Edit..Preferences..Advanced..Config Editor. Search for:
412 mailnews.send_plaintext_flowed
413 toggle it to make sure it is set to 'false'.
414
415 After that is done, you should be able to compose email as you
416 otherwise would (cut + paste, git-format-patch | git-imap-send, etc),
417 and the patches should not be mangled.
418
419 Approach #2 (external editor):
420
421 This recipe appears to work with the current [*1*] Thunderbird from Suse.
422
423 The following Thunderbird extensions are needed:
424 AboutConfig 0.5
425 http://aboutconfig.mozdev.org/
426 External Editor 0.7.2
427 http://globs.org/articles.php?lng=en&pg=8
428
429 1) Prepare the patch as a text file using your method of choice.
430
431 2) Before opening a compose window, use Edit->Account Settings to
432 uncheck the "Compose messages in HTML format" setting in the
433 "Composition & Addressing" panel of the account to be used to send the
434 patch. [*2*]
435
436 3) In the main Thunderbird window, _before_ you open the compose window
437 for the patch, use Tools->about:config to set the following to the
438 indicated values:
439 mailnews.send_plaintext_flowed => false
440 mailnews.wraplength => 0
441
442 4) Open a compose window and click the external editor icon.
443
444 5) In the external editor window, read in the patch file and exit the
445 editor normally.
446
447 6) Back in the compose window: Add whatever other text you wish to the
448 message, complete the addressing and subject fields, and press send.
449
450 7) Optionally, undo the about:config/account settings changes made in
451 steps 2 & 3.
452
453
454 [Footnotes]
455 *1* Version 1.0 (20041207) from the MozillaThunderbird-1.0-5 rpm of Suse
456 9.3 professional updates.
457
458 *2* It may be possible to do this with about:config and the following
459 settings but I haven't tried, yet.
460 mail.html_compose => false
461 mail.identity.default.compose_html => false
462 mail.identity.id?.compose_html => false
463
464 (Lukas Sandström)
465
466 There is a script in contrib/thunderbird-patch-inline which can help
467 you include patches with Thunderbird in an easy way. To use it, do the
468 steps above and then use the script as the external editor.
469
470 Gnus
471 ----
472
473 '|' in the *Summary* buffer can be used to pipe the current
474 message to an external program, and this is a handy way to drive
475 "git am". However, if the message is MIME encoded, what is
476 piped into the program is the representation you see in your
477 *Article* buffer after unwrapping MIME. This is often not what
478 you would want for two reasons. It tends to screw up non ASCII
479 characters (most notably in people's names), and also
480 whitespaces (fatal in patches). Running 'C-u g' to display the
481 message in raw form before using '|' to run the pipe can work
482 this problem around.
483
484
485 KMail
486 -----
487
488 This should help you to submit patches inline using KMail.
489
490 1) Prepare the patch as a text file.
491
492 2) Click on New Mail.
493
494 3) Go under "Options" in the Composer window and be sure that
495 "Word wrap" is not set.
496
497 4) Use Message -> Insert file... and insert the patch.
498
499 5) Back in the compose window: add whatever other text you wish to the
500 message, complete the addressing and subject fields, and press send.
501
502
503 Gmail
504 -----
505
506 GMail does not appear to have any way to turn off line wrapping in the web
507 interface, so this will mangle any emails that you send. You can however
508 use any IMAP email client to connect to the google imap server, and forward
509 the emails through that. Just make sure to disable line wrapping in that
510 email client. Alternatively, use "git send-email" instead.
511
512 Submitting properly formatted patches via Gmail is simple now that
513 IMAP support is available. First, edit your ~/.gitconfig to specify your
514 account settings:
515
516 [imap]
517 folder = "[Gmail]/Drafts"
518 host = imaps://imap.gmail.com
519 user = user@gmail.com
520 pass = p4ssw0rd
521 port = 993
522 sslverify = false
523
524 You might need to instead use: folder = "[Google Mail]/Drafts" if you get an error
525 that the "Folder doesn't exist".
526
527 Next, ensure that your Gmail settings are correct. In "Settings" the
528 "Use Unicode (UTF-8) encoding for outgoing messages" should be checked.
529
530 Once your commits are ready to send to the mailing list, run the following
531 command to send the patch emails to your Gmail Drafts folder.
532
533 $ git format-patch -M --stdout origin/master | git imap-send
534
535 Go to your Gmail account, open the Drafts folder, find the patch email, fill
536 in the To: and CC: fields and send away!
537