| blob | dd201848b2c9d8c36ddeae427014749b7ee5093d |
3 <html>
4 <head>
8 </head>
9 <body>
12 <ol>
15 <ol>
26 </ol></li>
28 <ol>
33 </ol></li>
34 </ol>
37 <!--=========================================================================-->
39 <!--=========================================================================-->
41 <p>This document contains the LLVM Developer Policy which defines the project's
42 policy towards developers and their contributions. The intent of this policy
43 is to eliminate miscommunication, rework, and confusion that might arise from
44 the distributed nature of LLVM's development. By stating the policy in clear
45 terms, we hope each developer can know ahead of time what to expect when
46 making LLVM contributions.</p>
49 <ol>
55 </ol>
57 <p>This policy is aimed at frequent contributors to LLVM. People interested in
58 contributing one-off patches can do so in an informal way by sending them to
59 the
61 mailing list</a> and engaging another developer to see it through the
62 process.</p>
63 </div>
65 <!--=========================================================================-->
67 <!--=========================================================================-->
69 <p>This section contains policies that pertain to frequent LLVM developers. We
71 routinely contribute to LLVM, but we expect more from frequent contributors
72 to keep the system as efficient as possible for everyone. Frequent LLVM
73 contributors are expected to meet the following requirements in order for
74 LLVM to maintain a high standard of quality.<p>
75 </div>
77 <!-- _______________________________________________________________________ -->
80 <p>Developers should stay informed by reading at least the
82 list. If you are doing anything more than just casual work on LLVM, it is
83 suggested that you also subscribe to the
85 list and pay attention to changes being made by others.</p>
87 <p>We recommend that active developers register an email account with
90 email list to keep track of bugs and enhancements occurring in LLVM.</p>
91 </div>
93 <!-- _______________________________________________________________________ -->
97 <p>When making a patch for review, the goal is to make it as easy for the
98 reviewer to read it as possible. As such, we recommend that you:</p>
100 <ol>
101 <li>Make your patch against the Subversion trunk, not a branch, and not an old
102 version of LLVM. This makes it easy to apply the patch.</li>
104 <li>Similarly, patches should be submitted soon after they are generated. Old
105 patches may not apply correctly if the underlying code changes between the
106 time the patch was created and the time it is applied.</li>
108 <li>Patches should be made with this command:
110 <pre>
111 svn diff
112 </pre>
113 </div>
115 the diff.</li>
117 <li>Patches should not include differences in generated code such as the code
120 </ol>
122 <p>When sending a patch to a mailing list, it is a good idea to send it as an
124 message. This ensures that your mailer will not mangle the patch when it
125 sends it (e.g. by making whitespace changes or by wrapping lines).</p>
132 attachment</tt>. Apple Mail gamely displays such a file inline, making it
133 difficult to work with for reviewers using that program.</p>
134 </div>
136 <!-- _______________________________________________________________________ -->
139 <p>LLVM has a code review policy. Code review is one way to increase the quality
140 of software. We generally follow these policies:</p>
142 <ol>
143 <li>All developers are required to have significant changes reviewed before
144 they are committed to the repository.</li>
146 <li>Code reviews are conducted by email, usually on the llvm-commits
147 list.</li>
149 <li>Code can be reviewed either before it is committed or after. We expect
150 major changes to be reviewed before being committed, but smaller changes
151 (or changes where the developer owns the component) can be reviewed after
152 commit.</li>
154 <li>The developer responsible for a code change is also responsible for making
155 all necessary review-related changes.</li>
157 <li>Code review can be an iterative process, which continues until the patch
158 is ready to be committed.</li>
159 </ol>
161 <p>Developers should participate in code reviews as both reviewers and
162 reviewees. If someone is kind enough to review your code, you should return
163 the favor for someone else. Note that anyone is welcome to review and give
164 feedback on a patch, but only people with Subversion write access can approve
165 it.</p>
166 </div>
168 <!-- _______________________________________________________________________ -->
172 <p>The LLVM Project relies on two features of its process to maintain rapid
173 development in addition to the high quality of its source base: the
174 combination of code review plus post-commit review for trusted maintainers.
175 Having both is a great way for the project to take advantage of the fact that
176 most people do the right thing most of the time, and only commit patches
177 without pre-commit review when they are confident they are right.</p>
179 <p>The trick to this is that the project has to guarantee that all patches that
180 are committed are reviewed after they go in: you don't want everyone to
181 assume someone else will review it, allowing the patch to go unreviewed. To
182 solve this problem, we have a notion of an 'owner' for a piece of the code.
183 The sole responsibility of a code owner is to ensure that a commit to their
184 area of the code is appropriately reviewed, either by themself or by someone
185 else. The current code owners are:</p>
187 <ol>
189 Windows codegen.</li>
196 </ol>
198 <p>Note that code ownership is completely different than reviewers: anyone can
199 review a piece of code, and we welcome code review from anyone who is
200 interested. Code owners are the "last line of defense" to guarantee that all
201 patches that are committed are actually reviewed.</p>
203 <p>Being a code owner is a somewhat unglamorous position, but it is incredibly
204 important for the ongoing success of the project. Because people get busy,
205 interests change, and unexpected things happen, code ownership is purely
206 opt-in, and anyone can choose to resign their "title" at any time. For now,
207 we do not have an official policy on how one gets elected to be a code
208 owner.</p>
209 </div>
211 <!-- _______________________________________________________________________ -->
214 <p>Developers are required to create test cases for any bugs fixed and any new
215 features added. Some tips for getting your testcase approved:</p>
217 <ol>
218 <li>All feature and regression test cases are added to the
221 details).</li>
224 language</a> unless the feature or regression being tested requires
225 another language (e.g. the bug being fixed or feature being implemented is
226 in the llvm-gcc C++ front-end, in which case it must be written in
227 C++).</li>
229 <li>Test cases, especially for regressions, should be reduced as much as
233 them short.</li>
234 </ol>
236 <p>Note that llvm/test is designed for regression and small feature tests
237 only. More extensive test cases (e.g., entire applications, benchmarks, etc)
239 for coverage (correctness, performance, etc) testing, not feature or
240 regression testing.</p>
241 </div>
243 <!-- _______________________________________________________________________ -->
246 <p>The minimum quality standards that any change must satisfy before being
247 committed to the main development branch are:</p>
249 <ol>
253 <li>Code must compile cleanly (no errors, no warnings) on at least one
254 platform.</li>
257 testcase</a> so we know if the fix/feature ever regresses in the
258 future.</li>
262 <li>The code must not cause regressions on a reasonable subset of llvm-test,
263 where "reasonable" depends on the contributor's judgement and the scope of
264 the change (more invasive changes require more testing). A reasonable
265 subset might be something like
267 </ol>
269 <p>Additionally, the committer is responsible for addressing any problems found
270 in the future that the change is responsible for. For example:</p>
272 <ul>
275 <li>The changes should not cause any correctness regressions in the
277 regressions.</li>
279 <li>The change set should not cause performance or correctness regressions for
280 the LLVM tools.</li>
282 <li>The changes should not cause performance or correctness regressions in
283 code compiled by LLVM on all applicable targets.</li>
287 </ul>
289 <p>We prefer for this to be handled before submission but understand that it
290 isn't possible to test all of this for every submission. Our build bots and
291 nightly testing infrastructure normally finds these problems. A good rule of
292 thumb is to check the nightly testers for regressions the day after your
293 change. Build bots will directly email you if a group of commits that
294 included yours caused a failure. You are expected to check the build bot
295 messages to see if they are your fault and, if so, fix the breakage.</p>
297 <p>Commits that violate these quality standards (e.g. are very broken) may be
298 reverted. This is necessary when the change blocks other developers from
299 making progress. The developer is welcome to re-commit the change after the
300 problem has been fixed.</p>
301 </div>
303 <!-- _______________________________________________________________________ -->
308 <p>We grant commit access to contributors with a track record of submitting high
309 quality patches. If you would like commit access, please send an email to
311 information:</p>
313 <ol>
316 <li>The full name and email address you want message to llvm-commits to come
320 Note that you don't ever tell us what your password is, you just give it
321 to us in an encrypted form. To get this, run "htpasswd" (a utility that
322 comes with apache) in crypt mode (often enabled with "-d"), or find a web
323 page that will do it for you.</li>
324 </ol>
326 <p>Once you've been granted commit access, you should be able to check out an
327 LLVM tree with an SVN URL of "https://username@llvm.org/..." instead of the
328 normal anonymous URL of "http://llvm.org/...". The first time you commit
329 you'll have to type in your password. Note that you may get a warning from
330 SVN about an untrusted key, you can ignore this. To verify that your commit
331 access works, please do a test commit (e.g. change a comment or add a blank
332 line). Your first commit to a repository may require the autogenerated email
333 to be approved by a mailing list. This is normal, and will be done when
334 the mailing list owner has time.</p>
338 <ol>
342 When approved you may commit it yourself.</li>
344 <li>You are allowed to commit patches without approval which you think are
345 obvious. This is clearly a subjective decision — we simply expect
346 you to use good judgement. Examples include: fixing build breakage,
347 reverting obviously broken patches, documentation/comment changes, any
348 other minor changes.</li>
350 <li>You are allowed to commit patches without approval to those portions of
351 LLVM that you have contributed or maintain (i.e., have been assigned
352 responsibility for), with the proviso that such commits must not break the
353 build. This is a "trust but verify" policy and commits of this nature are
354 reviewed after they are committed.</li>
356 <li>Multiple violations of these policies or a single egregious violation may
357 cause commit access to be revoked.</li>
358 </ol>
361 review</a> (either before or after they are committed, depending on the
362 nature of the change). You are encouraged to review other peoples' patches
363 as well, but you aren't required to.</p>
364 </div>
366 <!-- _______________________________________________________________________ -->
369 <p>When a developer begins a major new project with the aim of contributing it
370 back to LLVM, s/he should inform the community with an email to
372 email list, to the extent possible. The reason for this is to:
374 <ol>
377 <li>avoid duplication of effort by preventing multiple parties working on the
378 same thing and not knowing about it, and</li>
380 <li>ensure that any technical issues around the proposed work are discussed
381 and resolved before any significant work is done.</li>
382 </ol>
384 <p>The design of LLVM is carefully controlled to ensure that all the pieces fit
385 together well and are as consistent as possible. If you plan to make a major
386 change to the way LLVM works or want to add a major new extension, it is a
387 good idea to get consensus with the development community before you start
388 working on it.</p>
390 <p>Once the design of the new feature is finalized, the work itself should be
392 long-term development branch.</p>
393 </div>
395 <!-- _______________________________________________________________________ -->
397 </div>
399 <p>In the LLVM project, we do all significant changes as a series of incremental
400 patches. We have a strong dislike for huge changes or long-term development
401 branches. Long-term development branches have a number of drawbacks:</p>
403 <ol>
404 <li>Branches must have mainline merged into them periodically. If the branch
405 development and mainline development occur in the same pieces of code,
406 resolving merge conflicts can take a lot of time.</li>
410 <li>Huge changes (produced when a branch is merged back onto mainline) are
413 <li>Branches are not routinely tested by our nightly tester
414 infrastructure.</li>
416 <li>Changes developed as monolithic large changes often don't work until the
417 entire set of changes is done. Breaking it down into a set of smaller
418 changes increases the odds that any of the work will be committed to the
419 main repository.</li>
420 </ol>
422 <p>To address these problems, LLVM uses an incremental development style and we
423 require contributors to follow this practice when making a large/invasive
424 change. Some tips:</p>
426 <ul>
427 <li>Large/invasive changes usually have a number of secondary changes that are
428 required before the big change can be made (e.g. API cleanup, etc). These
429 sorts of changes can often be done before the major change is done,
430 independently of that work.</li>
432 <li>The remaining inter-related work should be decomposed into unrelated sets
433 of changes if possible. Once this is done, define the first increment and
434 get consensus on what the end goal of the change is.</li>
436 <li>Each change in the set can be stand alone (e.g. to fix a bug), or part of
437 a planned series of changes that works towards the development goal.</li>
439 <li>Each change should be kept as small as possible. This simplifies your work
440 (into a logical progression), simplifies code review and reduces the
441 chance that you will get negative feedback on the change. Small increments
442 also facilitate the maintenance of a high quality code base.</li>
444 <li>Often, an independent precursor to a big change is to add a new API and
445 slowly migrate clients to use the new API. Each change to use the new API
446 is often "obvious" and can be committed without review. Once the new API
447 is in place and used, it is much easier to replace the underlying
448 implementation of the API. This implementation change is logically
449 separate from the API change.</li>
450 </ul>
452 <p>If you are interested in making a large change, and this scares you, please
454 then ask about the best way to go about making the change.</p>
455 </div>
457 <!-- _______________________________________________________________________ -->
459 Changes</a></div>
461 <p>We believe in correct attribution of contributions to their contributors.
462 However, we do not want the source code to be littered with random
463 attributions "this code written by J. Random Hacker" (this is noisy and
464 distracting). In practice, the revision control system keeps a perfect
465 history of who changed what, and the CREDITS.txt file describes higher-level
466 contributions. If you commit a patch for someone else, please say "patch
470 </div>
472 <!--=========================================================================-->
475 </div>
476 <!--=========================================================================-->
479 <p>This section addresses the issues of copyright, license and patents for the
480 LLVM project. Currently, the University of Illinois is the LLVM copyright
481 holder and the terms of its license to LLVM users and developers is the
487 legal matters but does not provide legal advice. We are not lawyers, please
488 seek legal counsel from an attorney.</p>
489 </div>
490 </div>
492 <!-- _______________________________________________________________________ -->
495 <p>For consistency and ease of management, the project requires the copyright
496 for all LLVM software to be held by a single copyright holder: the University
497 of Illinois (UIUC).</p>
499 <p>Although UIUC may eventually reassign the copyright of the software to
500 another entity (e.g. a dedicated non-profit "LLVM Organization") the intent
501 for the project is to always have a single entity hold the copyrights to LLVM
502 at any given time.</p>
504 <p>We believe that having a single copyright holder is in the best interests of
505 all developers and users as it greatly reduces the managerial burden for any
506 kind of administrative or technical decisions about LLVM. The goal of the
509 </div>
511 <!-- _______________________________________________________________________ -->
514 <p>We intend to keep LLVM perpetually open source and to use a liberal open
515 source license. The current license is the
519 <ul>
524 <li>Binaries derived from LLVM must reproduce the copyright notice (e.g. in
525 an included readme file).</li>
530 </ul>
533 commercial products to be derived from LLVM</b> with few restrictions and
534 without a requirement for making any derived works also open source (i.e.
535 LLVM's license is not a "copyleft" license like the GPL). We suggest that you
537 if further clarification is needed.</p>
540 This means that anything "linked" into llvm-gcc must itself be compatible
541 with the GPL, and must be releasable under the terms of the GPL. This
542 implies that <b>any code linked into llvm-gcc and distributed to others may
543 be subject to the viral aspects of the GPL</b> (for example, a proprietary
544 code generator linked into llvm-gcc must be made available under the GPL).
545 This is not a problem for code already distributed under a more liberal
546 license (like the UIUC license), and does not affect code generated by
547 llvm-gcc. It may be a problem if you intend to base commercial development
548 on llvm-gcc without redistributing your source code.</p>
550 <p>We have no plans to change the license of LLVM. If you have questions or
551 comments about the license, please contact the
553 </div>
555 <!-- _______________________________________________________________________ -->
558 <p>To the best of our knowledge, LLVM does not infringe on any patents (we have
559 actually removed code from LLVM in the past that was found to infringe).
560 Having code in LLVM that infringes on patents would violate an important goal
561 of the project by making it hard or impossible to reuse the code for
562 arbitrary purposes (including commercial use).</p>
564 <p>When contributing code, we expect contributors to notify us of any potential
565 for patent-related trouble with their changes. If you or your employer own
566 the rights to a patent and would like to contribute code to LLVM that relies
567 on it, we require that the copyright owner sign an agreement that allows any
568 other user of LLVM to freely use your patent. Please contact
570 details.</p>
571 </div>
573 <!-- _______________________________________________________________________ -->
576 <p>With regards to the LLVM copyright and licensing, developers agree to assign
577 their copyrights to UIUC for any contribution made so that the entire
578 software base can be managed by a single copyright holder. This implies that
579 any contributions can be licensed under the license that the project
580 uses.</p>
582 <p>When contributing code, you also affirm that you are legally entitled to
583 grant this copyright, personally or on behalf of your employer. If the code
584 belongs to some other entity, please raise this issue with the oversight
585 group before the code is committed.</p>
586 </div>
588 <!-- *********************************************************************** -->
589 <hr>
590 <address>
595 Written by the
598 Last modified: $Date$
599 </address>
600 </body>
601 </html>