Renato Golin via llvm-dev
2020-Nov-02 19:58 UTC
[llvm-dev] Policy on support tiers, take 2
Ok, so after some feedback, here's an updated version. Separate thread as the previous got split. People seem to agree on the overall terms, but there was confusion on the difference between tier 2 and 3 as well as clarification on what projects go where. I have joined tiers 2 and 3 and made explicit the three criteria they fit into, with the requirements more formally explained. Please review the sub-project lists on the two tiers, I'm not so sure about them. Once we're happy with the "what", I'll send a review for a new doc so we can discuss the writing and format there (ignore it for now). Here it goes: *** Tier 1: the core compiler, officially supported by the community. Rationale: * Common code that supports most projects, upstream and downstream forks and forms the toolchain itself. * Includes everything we release on all architectures and OSs we have releases on. What: * LLVM itself, clang/tools, compiler-rt, libcxx/abi/unwind, lld, lldb, openmp, mlir. * Basically everything inside the mono-repo that is not in tier 2. * Builds on all first class citizen combinations of targets x OSs (incl. test-suite). * The CMake build infrastructure and release scripts. * Phabricator & Buildbot infrastructure. * The test-suite repository. Requirements: * Follow all core development policies on code quality, reviews, reverts, etc. * Noisy green buildbots, emailing all developers. * Most not be broken by changes in tier 2 (ex. if they require tier 1 changes). * Bitrot will downgrade areas to tier 2 or be removed, depending if a sub-community picks up support and has a timeline to fix. *** Tier 2: side projects that integrate with the compiler and that *should* work at all times. Rationale: * Code that is making its way into LLVM core (tier 1) via experimental/incubator roadmaps, or; * Code that isn't meant to be in LLVM core, but has a sub-community that maintains it long term, or; * Code that is making its way out of LLVM core (legacy) and that is a strong candidate for removal. What: * Experimental targets/options. * Experimental mono-repo projects (flang, libc, libclc, parallel-libs, polly, beduginfo-tests?, pstl?) * Incubator projects (circt, mlir-npcomp, etc). * Legacy tools (lnt). * Alternative build systems (gn, bazel). * Tool support (gdb scripts, editor configuration, helper scripts). Requirements: * Follow all core development policies on code quality, reviews, reverts, etc. * Infrastructure that only notify its sub-community. * Most not break tier 1, promptly reverting if it does, with discussions to be done offline before reapply. * Leaner policy on bots being red for longer, as long as the sub-community has a roadmap to fix. * Leaner policy on bitrot, as long as it doesn't impact tier 1 or other tier 2 projects. * Should be easy to remove (either separate path, or clear impact in code). * Must have a document making clear status, level of support and, if applicable, roadmap into tier 1 / out of LLVM. cheers, --renato -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20201102/36365f42/attachment.html>
Geoffrey Martin-Noble via llvm-dev
2020-Nov-02 20:27 UTC
[llvm-dev] Policy on support tiers, take 2
I'm not super familiar with all the various subprojects, but I'd be a little hesitant to put e.g. libc and flang in the same category as vim bindings or the gn build :-D I think I've had failing pre-merge checks from both flang and polly before, so at least in practice they don't seem to be following the guidelines you mention here. I don't know whether this means that they should just be moved up to tier 1, or whether they are actually less-supported than some of the more established projects. If their current behavior should change, then that seems like a bigger discussion. I also think it might be better to make this focus specifically on the monorepo. Incubator projects already have a clear policy and a lot of the fine points here only make sense in the context of the monorepo. e.g. bots with blamelists targeting only a subcommunity doesn't make much sense for a separate repo. If you committed the the mlir-npcomp repo I think you should get notified if you broke something :-D Apologies if the below falls into the category of "writing", but here are some additional thoughts: I also think that the distinction we previously had with tier 2 vs tier 3 was useful in terms of differentiating things that need quite a bit of promised support to be allowed in the monorepo because they're bigger, more complicated, and require constant maintenance (e.g build systems) vs things that can be checked in without much discussion because they are small, simple, and degrade gracefully (e.g. editor bindings). Maybe separate tiers isn't the right way to do that, but I think we should make that distinction clear. Like if someone wants to check in bindings for their editor of choice, a simple patch seems appropriate, whereas if someone (hypothetically ;-P) wants to propose a secondary build system it should at least be discussed on the mailing list. Maybe we can just include language in the description of tier 2, like: When adding components intending for tier 2 status, the level of discussion and support commitment required is proportional to the size and complexity of the component. For example: 1. editor bindings can be just be added as a patch following the normal review process 2. more complex components like a secondary build system should start as an RFC that details how this component will be supported 3. Experimental backends have an entire section on their introduction (link) On Mon, Nov 2, 2020 at 11:58 AM Renato Golin <rengolin at gmail.com> wrote:> Ok, so after some feedback, here's an updated version. Separate thread as > the previous got split. > > People seem to agree on the overall terms, but there was confusion on the > difference between tier 2 and 3 as well as clarification on what projects > go where. > > I have joined tiers 2 and 3 and made explicit the three criteria they fit > into, with the requirements more formally explained. > > Please review the sub-project lists on the two tiers, I'm not so sure > about them. > > Once we're happy with the "what", I'll send a review for a new doc so we > can discuss the writing and format there (ignore it for now). > > Here it goes: > > *** Tier 1: the core compiler, officially supported by the community. > > Rationale: > * Common code that supports most projects, upstream and downstream forks > and forms the toolchain itself. > * Includes everything we release on all architectures and OSs we have > releases on. > > What: > * LLVM itself, clang/tools, compiler-rt, libcxx/abi/unwind, lld, lldb, > openmp, mlir. > * Basically everything inside the mono-repo that is not in tier 2. > * Builds on all first class citizen combinations of targets x OSs (incl. > test-suite). > * The CMake build infrastructure and release scripts. > * Phabricator & Buildbot infrastructure. > * The test-suite repository. > > Requirements: > * Follow all core development policies on code quality, reviews, reverts, > etc. > * Noisy green buildbots, emailing all developers. > * Most not be broken by changes in tier 2 (ex. if they require tier 1 > changes). > * Bitrot will downgrade areas to tier 2 or be removed, depending if a > sub-community picks up support and has a timeline to fix. > > *** Tier 2: side projects that integrate with the compiler and that > *should* work at all times. > > Rationale: > * Code that is making its way into LLVM core (tier 1) via > experimental/incubator roadmaps, or; > * Code that isn't meant to be in LLVM core, but has a sub-community that > maintains it long term, or; > * Code that is making its way out of LLVM core (legacy) and that is a > strong candidate for removal. > > What: > * Experimental targets/options. > * Experimental mono-repo projects (flang, libc, libclc, parallel-libs, > polly, beduginfo-tests?, pstl?) > * Incubator projects (circt, mlir-npcomp, etc). > * Legacy tools (lnt). > * Alternative build systems (gn, bazel). > * Tool support (gdb scripts, editor configuration, helper scripts). > > Requirements: > * Follow all core development policies on code quality, reviews, reverts, > etc. > * Infrastructure that only notify its sub-community. > * Most not break tier 1, promptly reverting if it does, with discussions > to be done offline before reapply. > * Leaner policy on bots being red for longer, as long as the > sub-community has a roadmap to fix. > * Leaner policy on bitrot, as long as it doesn't impact tier 1 or other > tier 2 projects. > * Should be easy to remove (either separate path, or clear impact in > code). > * Must have a document making clear status, level of support and, if > applicable, roadmap into tier 1 / out of LLVM. > > > cheers, > --renato >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20201102/6f395b76/attachment.html> -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 3992 bytes Desc: S/MIME Cryptographic Signature URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20201102/6f395b76/attachment.bin>
Eric Christopher via llvm-dev
2020-Nov-02 20:32 UTC
[llvm-dev] Policy on support tiers, take 2
I think I'd work at a little higher level on the things. Explicit lists are hard to maintain and, in general, I think we're looking at guidelines rather than hard lists. -eric On Mon, Nov 2, 2020 at 3:27 PM Geoffrey Martin-Noble <gcmn at google.com> wrote:> I'm not super familiar with all the various subprojects, but I'd be a > little hesitant to put e.g. libc and flang in the same category as vim > bindings or the gn build :-D I think I've had failing pre-merge checks from > both flang and polly before, so at least in practice they don't seem to be > following the guidelines you mention here. I don't know whether this means > that they should just be moved up to tier 1, or whether they are actually > less-supported than some of the more established projects. If their current > behavior should change, then that seems like a bigger discussion. > > I also think it might be better to make this focus specifically on the > monorepo. Incubator projects already have a clear policy and a lot of the > fine points here only make sense in the context of the monorepo. e.g. bots > with blamelists targeting only a subcommunity doesn't make much sense for a > separate repo. If you committed the the mlir-npcomp repo I think you should > get notified if you broke something :-D > > Apologies if the below falls into the category of "writing", but here are > some additional thoughts: > > I also think that the distinction we previously had with tier 2 vs tier 3 > was useful in terms of differentiating things that need quite a bit of > promised support to be allowed in the monorepo because they're bigger, more > complicated, and require constant maintenance (e.g build systems) vs things > that can be checked in without much discussion because they are small, > simple, and degrade gracefully (e.g. editor bindings). Maybe separate tiers > isn't the right way to do that, but I think we should make that distinction > clear. Like if someone wants to check in bindings for their editor of > choice, a simple patch seems appropriate, whereas if someone > (hypothetically ;-P) wants to propose a secondary build system it should at > least be discussed on the mailing list. Maybe we can just include language > in the description of tier 2, like: > > When adding components intending for tier 2 status, the level of > discussion and support commitment required is proportional to the size and > complexity of the component. For example: > 1. editor bindings can be just be added as a patch following the normal > review process > 2. more complex components like a secondary build system should start as > an RFC that details how this component will be supported > 3. Experimental backends have an entire section on their introduction > (link) > > On Mon, Nov 2, 2020 at 11:58 AM Renato Golin <rengolin at gmail.com> wrote: > >> Ok, so after some feedback, here's an updated version. Separate thread as >> the previous got split. >> >> People seem to agree on the overall terms, but there was confusion on the >> difference between tier 2 and 3 as well as clarification on what projects >> go where. >> >> I have joined tiers 2 and 3 and made explicit the three criteria they fit >> into, with the requirements more formally explained. >> >> Please review the sub-project lists on the two tiers, I'm not so sure >> about them. >> >> Once we're happy with the "what", I'll send a review for a new doc so we >> can discuss the writing and format there (ignore it for now). >> >> Here it goes: >> >> *** Tier 1: the core compiler, officially supported by the community. >> >> Rationale: >> * Common code that supports most projects, upstream and downstream forks >> and forms the toolchain itself. >> * Includes everything we release on all architectures and OSs we have >> releases on. >> >> What: >> * LLVM itself, clang/tools, compiler-rt, libcxx/abi/unwind, lld, lldb, >> openmp, mlir. >> * Basically everything inside the mono-repo that is not in tier 2. >> * Builds on all first class citizen combinations of targets x OSs (incl. >> test-suite). >> * The CMake build infrastructure and release scripts. >> * Phabricator & Buildbot infrastructure. >> * The test-suite repository. >> >> Requirements: >> * Follow all core development policies on code quality, reviews, >> reverts, etc. >> * Noisy green buildbots, emailing all developers. >> * Most not be broken by changes in tier 2 (ex. if they require tier 1 >> changes). >> * Bitrot will downgrade areas to tier 2 or be removed, depending if a >> sub-community picks up support and has a timeline to fix. >> >> *** Tier 2: side projects that integrate with the compiler and that >> *should* work at all times. >> >> Rationale: >> * Code that is making its way into LLVM core (tier 1) via >> experimental/incubator roadmaps, or; >> * Code that isn't meant to be in LLVM core, but has a sub-community that >> maintains it long term, or; >> * Code that is making its way out of LLVM core (legacy) and that is a >> strong candidate for removal. >> >> What: >> * Experimental targets/options. >> * Experimental mono-repo projects (flang, libc, libclc, parallel-libs, >> polly, beduginfo-tests?, pstl?) >> * Incubator projects (circt, mlir-npcomp, etc). >> * Legacy tools (lnt). >> * Alternative build systems (gn, bazel). >> * Tool support (gdb scripts, editor configuration, helper scripts). >> >> Requirements: >> * Follow all core development policies on code quality, reviews, >> reverts, etc. >> * Infrastructure that only notify its sub-community. >> * Most not break tier 1, promptly reverting if it does, with discussions >> to be done offline before reapply. >> * Leaner policy on bots being red for longer, as long as the >> sub-community has a roadmap to fix. >> * Leaner policy on bitrot, as long as it doesn't impact tier 1 or other >> tier 2 projects. >> * Should be easy to remove (either separate path, or clear impact in >> code). >> * Must have a document making clear status, level of support and, if >> applicable, roadmap into tier 1 / out of LLVM. >> >> >> cheers, >> --renato >> >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20201102/21af9750/attachment.html>
Christian Kühnel via llvm-dev
2020-Nov-03 07:39 UTC
[llvm-dev] Policy on support tiers, take 2
Hi Renato, Thank you for working on this! When looking at the topic from the pre-merge testing perspective 2 questions came to my mind: 1) How should pre-merge testing behave differently on tier 1 vs. 2 projects? Do we want the builds/tests to pass on broken tier 2 projects? 2) How can we automatically detect the tiers of the projects in the pre-merge build scripts? If we want to treat them differently we somehow need to be able to distinguish these in the repo. Best, Christian On Mon, Nov 2, 2020 at 8:58 PM Renato Golin via llvm-dev < llvm-dev at lists.llvm.org> wrote:> Ok, so after some feedback, here's an updated version. Separate thread as > the previous got split. > > People seem to agree on the overall terms, but there was confusion on the > difference between tier 2 and 3 as well as clarification on what projects > go where. > > I have joined tiers 2 and 3 and made explicit the three criteria they fit > into, with the requirements more formally explained. > > Please review the sub-project lists on the two tiers, I'm not so sure > about them. > > Once we're happy with the "what", I'll send a review for a new doc so we > can discuss the writing and format there (ignore it for now). > > Here it goes: > > *** Tier 1: the core compiler, officially supported by the community. > > Rationale: > * Common code that supports most projects, upstream and downstream forks > and forms the toolchain itself. > * Includes everything we release on all architectures and OSs we have > releases on. > > What: > * LLVM itself, clang/tools, compiler-rt, libcxx/abi/unwind, lld, lldb, > openmp, mlir. > * Basically everything inside the mono-repo that is not in tier 2. > * Builds on all first class citizen combinations of targets x OSs (incl. > test-suite). > * The CMake build infrastructure and release scripts. > * Phabricator & Buildbot infrastructure. > * The test-suite repository. > > Requirements: > * Follow all core development policies on code quality, reviews, reverts, > etc. > * Noisy green buildbots, emailing all developers. > * Most not be broken by changes in tier 2 (ex. if they require tier 1 > changes). > * Bitrot will downgrade areas to tier 2 or be removed, depending if a > sub-community picks up support and has a timeline to fix. > > *** Tier 2: side projects that integrate with the compiler and that > *should* work at all times. > > Rationale: > * Code that is making its way into LLVM core (tier 1) via > experimental/incubator roadmaps, or; > * Code that isn't meant to be in LLVM core, but has a sub-community that > maintains it long term, or; > * Code that is making its way out of LLVM core (legacy) and that is a > strong candidate for removal. > > What: > * Experimental targets/options. > * Experimental mono-repo projects (flang, libc, libclc, parallel-libs, > polly, beduginfo-tests?, pstl?) > * Incubator projects (circt, mlir-npcomp, etc). > * Legacy tools (lnt). > * Alternative build systems (gn, bazel). > * Tool support (gdb scripts, editor configuration, helper scripts). > > Requirements: > * Follow all core development policies on code quality, reviews, reverts, > etc. > * Infrastructure that only notify its sub-community. > * Most not break tier 1, promptly reverting if it does, with discussions > to be done offline before reapply. > * Leaner policy on bots being red for longer, as long as the > sub-community has a roadmap to fix. > * Leaner policy on bitrot, as long as it doesn't impact tier 1 or other > tier 2 projects. > * Should be easy to remove (either separate path, or clear impact in > code). > * Must have a document making clear status, level of support and, if > applicable, roadmap into tier 1 / out of LLVM. > > > cheers, > --renato > _______________________________________________ > LLVM Developers mailing list > llvm-dev at lists.llvm.org > https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20201103/d2a27855/attachment.html>
Renato Golin via llvm-dev
2020-Nov-03 09:02 UTC
[llvm-dev] Policy on support tiers, take 2
On Tue, 3 Nov 2020 at 07:40, Christian Kühnel <kuhnel at google.com> wrote:> 1) How should pre-merge testing behave differently on tier 1 vs. 2 > projects? > Do we want the builds/tests to pass on broken tier 2 projects? >Hi Christian, My view is that we shouldn't change how we test, either pre merge or post merge. If we have official tests for something already, those must pass. The idea of tiers is to reduce the burden of maintenance of sections of the code for both the sub-community that cares about it and the rest of the community that doesn't. So, whatever we already have a pre-merge test for, is IMO tier 1. If they are causing grief and are not seen to be core components by the large LLVM community, then as we "demote" them to tier 2, we also remove them from the noisy testing (on both pre/post merge). Tier 2 testing is the responsibility of the sub-community that cares about it. If they want pre-merge tests for those, they're responsible for making sure the fixes are proposed as soon as possible for the patches in flight. None of those pre-commit tests should warn the rest of the community, or it would get hard to know which bot results we should "respect", especially for newcomers. 2) How can we automatically detect the tiers of the projects in the> pre-merge build scripts? If we want to treat them differently we somehow > need to be able to distinguish these in the repo. >It should be very simple to do that, but my second take had the mistake of listing projects by name. Others have suggested we don't do that and let the boundaries be more subtle. I very much like that. The idea is that there are concepts of tiers, and what goes in/out will change with time, depending on how big a sub-community a particular piece of code has, or on its experimental status, or if/when incubator projects move into the monorepo. Initially, everything outside of the mono-repo is tier 2, which matches the incubator policy quite well. Inside the monorepo, there are a few rules of what's tier 2 (experimental, non-CMake build systems, editor/tools config, non-essential scripts), with everything else in tier 1. With the requirement of "being tested with a noisy bot" in tier 1, everything that is already noisy-tested now is in tier 1. If we want to demote something, we need to be explicit about it and allow the creation of a sub-community to form. Makes sense? cheers, --renato -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20201103/bb915781/attachment.html>