13 This document contains the LLVM Developer Policy which defines the project's
14 policy towards developers and their contributions. The intent of this policy is
15 to eliminate miscommunication, rework, and confusion that might arise from the
16 distributed nature of LLVM's development. By stating the policy in clear terms,
17 we hope each developer can know ahead of time what to expect when making LLVM
18 contributions. This policy covers all llvm.org subprojects, including Clang,
21 This policy is also designed to accomplish the following objectives:
23 #. Attract both users and developers to the LLVM project.
25 #. Make life as simple and easy for contributors as possible.
27 #. Keep the top of Subversion trees as stable as possible.
29 #. Establish awareness of the project's `copyright, license, and patent
30 policies`_ with contributors to the project.
32 This policy is aimed at frequent contributors to LLVM. People interested in
33 contributing one-off patches can do so in an informal way by sending them to the
34 `llvm-commits mailing list
35 <http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>`_ and engaging another
36 developer to see it through the process.
41 This section contains policies that pertain to frequent LLVM developers. We
42 always welcome `one-off patches`_ from people who do not routinely contribute to
43 LLVM, but we expect more from frequent contributors to keep the system as
44 efficient as possible for everyone. Frequent LLVM contributors are expected to
45 meet the following requirements in order for LLVM to maintain a high standard of
51 Developers should stay informed by reading at least the "dev" mailing list for
52 the projects you are interested in, such as `llvmdev
53 <http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev>`_ for LLVM, `cfe-dev
54 <http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev>`_ for Clang, or `lldb-dev
55 <http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev>`_ for LLDB. If you are
56 doing anything more than just casual work on LLVM, it is suggested that you also
57 subscribe to the "commits" mailing list for the subproject you're interested in,
59 <http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>`_, `cfe-commits
60 <http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits>`_, or `lldb-commits
61 <http://lists.cs.uiuc.edu/mailman/listinfo/lldb-commits>`_. Reading the
62 "commits" list and paying attention to changes being made by others is a good
63 way to see what other people are interested in and watching the flow of the
66 We recommend that active developers register an email account with `LLVM
67 Bugzilla <http://llvm.org/bugs/>`_ and preferably subscribe to the `llvm-bugs
68 <http://lists.cs.uiuc.edu/mailman/listinfo/llvmbugs>`_ email list to keep track
69 of bugs and enhancements occurring in LLVM. We really appreciate people who are
70 proactive at catching incoming bugs in their components and dealing with them
79 When making a patch for review, the goal is to make it as easy for the reviewer
80 to read it as possible. As such, we recommend that you:
82 #. Make your patch against the Subversion trunk, not a branch, and not an old
83 version of LLVM. This makes it easy to apply the patch. For information on
84 how to check out SVN trunk, please see the `Getting Started
85 Guide <GettingStarted.html#checkout>`_.
87 #. Similarly, patches should be submitted soon after they are generated. Old
88 patches may not apply correctly if the underlying code changes between the
89 time the patch was created and the time it is applied.
91 #. Patches should be made with ``svn diff``, or similar. If you use a
92 different tool, make sure it uses the ``diff -u`` format and that it
93 doesn't contain clutter which makes it hard to read.
95 #. If you are modifying generated files, such as the top-level ``configure``
96 script, please separate out those changes into a separate patch from the rest
99 When sending a patch to a mailing list, it is a good idea to send it as an
100 *attachment* to the message, not embedded into the text of the message. This
101 ensures that your mailer will not mangle the patch when it sends it (e.g. by
102 making whitespace changes or by wrapping lines).
104 *For Thunderbird users:* Before submitting a patch, please open *Preferences >
105 Advanced > General > Config Editor*, find the key
106 ``mail.content_disposition_type``, and set its value to ``1``. Without this
107 setting, Thunderbird sends your attachment using ``Content-Disposition: inline``
108 rather than ``Content-Disposition: attachment``. Apple Mail gamely displays such
109 a file inline, making it difficult to work with for reviewers using that
117 LLVM has a code review policy. Code review is one way to increase the quality of
118 software. We generally follow these policies:
120 #. All developers are required to have significant changes reviewed before they
121 are committed to the repository.
123 #. Code reviews are conducted by email, usually on the llvm-commits list.
125 #. Code can be reviewed either before it is committed or after. We expect major
126 changes to be reviewed before being committed, but smaller changes (or
127 changes where the developer owns the component) can be reviewed after commit.
129 #. The developer responsible for a code change is also responsible for making
130 all necessary review-related changes.
132 #. Code review can be an iterative process, which continues until the patch is
133 ready to be committed.
135 Developers should participate in code reviews as both reviewers and
136 reviewees. If someone is kind enough to review your code, you should return the
137 favor for someone else. Note that anyone is welcome to review and give feedback
138 on a patch, but only people with Subversion write access can approve it.
143 The LLVM Project relies on two features of its process to maintain rapid
144 development in addition to the high quality of its source base: the combination
145 of code review plus post-commit review for trusted maintainers. Having both is
146 a great way for the project to take advantage of the fact that most people do
147 the right thing most of the time, and only commit patches without pre-commit
148 review when they are confident they are right.
150 The trick to this is that the project has to guarantee that all patches that are
151 committed are reviewed after they go in: you don't want everyone to assume
152 someone else will review it, allowing the patch to go unreviewed. To solve this
153 problem, we have a notion of an 'owner' for a piece of the code. The sole
154 responsibility of a code owner is to ensure that a commit to their area of the
155 code is appropriately reviewed, either by themself or by someone else. The
156 current code owners are:
158 * **Evan Cheng**: Code generator and all targets
160 * **Greg Clayton**: LLDB
162 * **Doug Gregor**: Clang Frontend Libraries
164 * **Howard Hinnant**: libc++
166 * **Anton Korobeynikov**: Exception handling, debug information, and Windows
169 * **Ted Kremenek**: Clang Static Analyzer
171 * **Chris Lattner**: Everything not covered by someone else
173 * **John McCall**: Clang LLVM IR generation
175 * **Jakob Olesen**: Register allocators and TableGen
177 * **Duncan Sands**: dragonegg and llvm-gcc 4.2
179 * **Peter Collingbourne**: libclc
181 * **Tobias Grosser**: polly
183 Note that code ownership is completely different than reviewers: anyone can
184 review a piece of code, and we welcome code review from anyone who is
185 interested. Code owners are the "last line of defense" to guarantee that all
186 patches that are committed are actually reviewed.
188 Being a code owner is a somewhat unglamorous position, but it is incredibly
189 important for the ongoing success of the project. Because people get busy,
190 interests change, and unexpected things happen, code ownership is purely opt-in,
191 and anyone can choose to resign their "title" at any time. For now, we do not
192 have an official policy on how one gets elected to be a code owner.
194 .. _include a testcase:
199 Developers are required to create test cases for any bugs fixed and any new
200 features added. Some tips for getting your testcase approved:
202 * All feature and regression test cases are added to the ``llvm/test``
203 directory. The appropriate sub-directory should be selected (see the `Testing
204 Guide <TestingGuide.html>`_ for details).
206 * Test cases should be written in `LLVM assembly language <LangRef.html>`_
207 unless the feature or regression being tested requires another language
208 (e.g. the bug being fixed or feature being implemented is in the llvm-gcc C++
209 front-end, in which case it must be written in C++).
211 * Test cases, especially for regressions, should be reduced as much as possible,
212 by `bugpoint <Bugpoint.html>`_ or manually. It is unacceptable to place an
213 entire failing program into ``llvm/test`` as this creates a *time-to-test*
214 burden on all developers. Please keep them short.
216 Note that llvm/test and clang/test are designed for regression and small feature
217 tests only. More extensive test cases (e.g., entire applications, benchmarks,
218 etc) should be added to the ``llvm-test`` test suite. The llvm-test suite is
219 for coverage (correctness, performance, etc) testing, not feature or regression
225 The minimum quality standards that any change must satisfy before being
226 committed to the main development branch are:
228 #. Code must adhere to the `LLVM Coding Standards <CodingStandards.html>`_.
230 #. Code must compile cleanly (no errors, no warnings) on at least one platform.
232 #. Bug fixes and new features should `include a testcase`_ so we know if the
233 fix/feature ever regresses in the future.
235 #. Code must pass the ``llvm/test`` test suite.
237 #. The code must not cause regressions on a reasonable subset of llvm-test,
238 where "reasonable" depends on the contributor's judgement and the scope of
239 the change (more invasive changes require more testing). A reasonable subset
240 might be something like "``llvm-test/MultiSource/Benchmarks``".
242 Additionally, the committer is responsible for addressing any problems found in
243 the future that the change is responsible for. For example:
245 * The code should compile cleanly on all supported platforms.
247 * The changes should not cause any correctness regressions in the ``llvm-test``
248 suite and must not cause any major performance regressions.
250 * The change set should not cause performance or correctness regressions for the
253 * The changes should not cause performance or correctness regressions in code
254 compiled by LLVM on all applicable targets.
256 * You are expected to address any `Bugzilla bugs <http://llvm.org/bugs/>`_ that
257 result from your change.
259 We prefer for this to be handled before submission but understand that it isn't
260 possible to test all of this for every submission. Our build bots and nightly
261 testing infrastructure normally finds these problems. A good rule of thumb is
262 to check the nightly testers for regressions the day after your change. Build
263 bots will directly email you if a group of commits that included yours caused a
264 failure. You are expected to check the build bot messages to see if they are
265 your fault and, if so, fix the breakage.
267 Commits that violate these quality standards (e.g. are very broken) may be
268 reverted. This is necessary when the change blocks other developers from making
269 progress. The developer is welcome to re-commit the change after the problem has
272 Obtaining Commit Access
273 -----------------------
275 We grant commit access to contributors with a track record of submitting high
276 quality patches. If you would like commit access, please send an email to
277 `Chris <mailto:sabre@nondot.org>`_ with the following information:
279 #. The user name you want to commit with, e.g. "hacker".
281 #. The full name and email address you want message to llvm-commits to come
282 from, e.g. "J. Random Hacker <hacker@yoyodyne.com>".
284 #. A "password hash" of the password you want to use, e.g. "``2ACR96qjUqsyM``".
285 Note that you don't ever tell us what your password is, you just give it to
286 us in an encrypted form. To get this, run "``htpasswd``" (a utility that
287 comes with apache) in crypt mode (often enabled with "``-d``"), or find a web
288 page that will do it for you.
290 Once you've been granted commit access, you should be able to check out an LLVM
291 tree with an SVN URL of "https://username@llvm.org/..." instead of the normal
292 anonymous URL of "http://llvm.org/...". The first time you commit you'll have
293 to type in your password. Note that you may get a warning from SVN about an
294 untrusted key, you can ignore this. To verify that your commit access works,
295 please do a test commit (e.g. change a comment or add a blank line). Your first
296 commit to a repository may require the autogenerated email to be approved by a
297 mailing list. This is normal, and will be done when the mailing list owner has
300 If you have recently been granted commit access, these policies apply:
302 #. You are granted *commit-after-approval* to all parts of LLVM. To get
303 approval, submit a `patch`_ to `llvm-commits
304 <http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits>`_. When approved
305 you may commit it yourself.</li>
307 #. You are allowed to commit patches without approval which you think are
308 obvious. This is clearly a subjective decision --- we simply expect you to
309 use good judgement. Examples include: fixing build breakage, reverting
310 obviously broken patches, documentation/comment changes, any other minor
313 #. You are allowed to commit patches without approval to those portions of LLVM
314 that you have contributed or maintain (i.e., have been assigned
315 responsibility for), with the proviso that such commits must not break the
316 build. This is a "trust but verify" policy and commits of this nature are
317 reviewed after they are committed.
319 #. Multiple violations of these policies or a single egregious violation may
320 cause commit access to be revoked.
322 In any case, your changes are still subject to `code review`_ (either before or
323 after they are committed, depending on the nature of the change). You are
324 encouraged to review other peoples' patches as well, but you aren't required
327 .. _discuss the change/gather consensus:
329 Making a Major Change
330 ---------------------
332 When a developer begins a major new project with the aim of contributing it back
333 to LLVM, s/he should inform the community with an email to the `llvmdev
334 <http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev>`_ email list, to the extent
335 possible. The reason for this is to:
337 #. keep the community informed about future changes to LLVM,
339 #. avoid duplication of effort by preventing multiple parties working on the
340 same thing and not knowing about it, and
342 #. ensure that any technical issues around the proposed work are discussed and
343 resolved before any significant work is done.
345 The design of LLVM is carefully controlled to ensure that all the pieces fit
346 together well and are as consistent as possible. If you plan to make a major
347 change to the way LLVM works or want to add a major new extension, it is a good
348 idea to get consensus with the development community before you start working on
351 Once the design of the new feature is finalized, the work itself should be done
352 as a series of `incremental changes`_, not as a long-term development branch.
354 .. _incremental changes:
356 Incremental Development
357 -----------------------
359 In the LLVM project, we do all significant changes as a series of incremental
360 patches. We have a strong dislike for huge changes or long-term development
361 branches. Long-term development branches have a number of drawbacks:
363 #. Branches must have mainline merged into them periodically. If the branch
364 development and mainline development occur in the same pieces of code,
365 resolving merge conflicts can take a lot of time.
367 #. Other people in the community tend to ignore work on branches.
369 #. Huge changes (produced when a branch is merged back onto mainline) are
370 extremely difficult to `code review`_.
372 #. Branches are not routinely tested by our nightly tester infrastructure.
374 #. Changes developed as monolithic large changes often don't work until the
375 entire set of changes is done. Breaking it down into a set of smaller
376 changes increases the odds that any of the work will be committed to the main
379 To address these problems, LLVM uses an incremental development style and we
380 require contributors to follow this practice when making a large/invasive
383 * Large/invasive changes usually have a number of secondary changes that are
384 required before the big change can be made (e.g. API cleanup, etc). These
385 sorts of changes can often be done before the major change is done,
386 independently of that work.
388 * The remaining inter-related work should be decomposed into unrelated sets of
389 changes if possible. Once this is done, define the first increment and get
390 consensus on what the end goal of the change is.
392 * Each change in the set can be stand alone (e.g. to fix a bug), or part of a
393 planned series of changes that works towards the development goal.
395 * Each change should be kept as small as possible. This simplifies your work
396 (into a logical progression), simplifies code review and reduces the chance
397 that you will get negative feedback on the change. Small increments also
398 facilitate the maintenance of a high quality code base.
400 * Often, an independent precursor to a big change is to add a new API and slowly
401 migrate clients to use the new API. Each change to use the new API is often
402 "obvious" and can be committed without review. Once the new API is in place
403 and used, it is much easier to replace the underlying implementation of the
404 API. This implementation change is logically separate from the API
407 If you are interested in making a large change, and this scares you, please make
408 sure to first `discuss the change/gather consensus`_ then ask about the best way
409 to go about making the change.
411 Attribution of Changes
412 ----------------------
414 We believe in correct attribution of contributions to their contributors.
415 However, we do not want the source code to be littered with random attributions
416 "this code written by J. Random Hacker" (this is noisy and distracting). In
417 practice, the revision control system keeps a perfect history of who changed
418 what, and the CREDITS.txt file describes higher-level contributions. If you
419 commit a patch for someone else, please say "patch contributed by J. Random
420 Hacker!" in the commit message.
422 Overall, please do not add contributor names to the source code.
424 .. _copyright, license, and patent policies:
426 Copyright, License, and Patents
427 ===============================
431 This section deals with legal matters but does not provide legal advice. We
432 are not lawyers --- please seek legal counsel from an attorney.
434 This section addresses the issues of copyright, license and patents for the LLVM
435 project. The copyright for the code is held by the individual contributors of
436 the code and the terms of its license to LLVM users and developers is the
437 `University of Illinois/NCSA Open Source License
438 <http://www.opensource.org/licenses/UoI-NCSA.php>`_ (with portions dual licensed
439 under the `MIT License <http://www.opensource.org/licenses/mit-license.php>`_,
440 see below). As contributor to the LLVM project, you agree to allow any
441 contributions to the project to licensed under these terms.
446 The LLVM project does not require copyright assignments, which means that the
447 copyright for the code in the project is held by its respective contributors who
448 have each agreed to release their contributed code under the terms of the `LLVM
451 An implication of this is that the LLVM license is unlikely to ever change:
452 changing it would require tracking down all the contributors to LLVM and getting
453 them to agree that a license change is acceptable for their contribution. Since
454 there are no plans to change the license, this is not a cause for concern.
456 As a contributor to the project, this means that you (or your company) retain
457 ownership of the code you contribute, that it cannot be used in a way that
458 contradicts the license (which is a liberal BSD-style license), and that the
459 license for your contributions won't change without your approval in the
467 We intend to keep LLVM perpetually open source and to use a liberal open source
468 license. **As a contributor to the project, you agree that any contributions be
469 licensed under the terms of the corresponding subproject.** All of the code in
470 LLVM is available under the `University of Illinois/NCSA Open Source License
471 <http://www.opensource.org/licenses/UoI-NCSA.php>`_, which boils down to
474 * You can freely distribute LLVM.
475 * You must retain the copyright notice if you redistribute LLVM.
476 * Binaries derived from LLVM must reproduce the copyright notice (e.g. in an
477 included readme file).
478 * You can't use our names to promote your LLVM derived products.
479 * There's no warranty on LLVM at all.
481 We believe this fosters the widest adoption of LLVM because it **allows
482 commercial products to be derived from LLVM** with few restrictions and without
483 a requirement for making any derived works also open source (i.e. LLVM's
484 license is not a "copyleft" license like the GPL). We suggest that you read the
485 `License <http://www.opensource.org/licenses/UoI-NCSA.php>`_ if further
486 clarification is needed.
488 In addition to the UIUC license, the runtime library components of LLVM
489 (**compiler_rt, libc++, and libclc**) are also licensed under the `MIT License
490 <http://www.opensource.org/licenses/mit-license.php>`_, which does not contain
491 the binary redistribution clause. As a user of these runtime libraries, it
492 means that you can choose to use the code under either license (and thus don't
493 need the binary redistribution clause), and as a contributor to the code that
494 you agree that any contributions to these libraries be licensed under both
495 licenses. We feel that this is important for runtime libraries, because they
496 are implicitly linked into applications and therefore should not subject those
497 applications to the binary redistribution clause. This also means that it is ok
498 to move code from (e.g.) libc++ to the LLVM core without concern, but that code
499 cannot be moved from the LLVM core to libc++ without the copyright owner's
502 Note that the LLVM Project does distribute llvm-gcc and dragonegg, **which are
503 GPL.** This means that anything "linked" into llvm-gcc must itself be compatible
504 with the GPL, and must be releasable under the terms of the GPL. This implies
505 that **any code linked into llvm-gcc and distributed to others may be subject to
506 the viral aspects of the GPL** (for example, a proprietary code generator linked
507 into llvm-gcc must be made available under the GPL). This is not a problem for
508 code already distributed under a more liberal license (like the UIUC license),
509 and GPL-containing subprojects are kept in separate SVN repositories whose
510 LICENSE.txt files specifically indicate that they contain GPL code.
512 We have no plans to change the license of LLVM. If you have questions or
513 comments about the license, please contact the `LLVM Developer's Mailing
514 List <mailto:llvmdev@cs.uiuc.edu>`_.
519 To the best of our knowledge, LLVM does not infringe on any patents (we have
520 actually removed code from LLVM in the past that was found to infringe). Having
521 code in LLVM that infringes on patents would violate an important goal of the
522 project by making it hard or impossible to reuse the code for arbitrary purposes
523 (including commercial use).
525 When contributing code, we expect contributors to notify us of any potential for
526 patent-related trouble with their changes (including from third parties). If
527 you or your employer own the rights to a patent and would like to contribute
528 code to LLVM that relies on it, we require that the copyright owner sign an
529 agreement that allows any other user of LLVM to freely use your patent. Please
530 contact the `oversight group <mailto:llvm-oversight@cs.uiuc.edu>`_ for more