[llvm-dev] RFC: Contributing Bazel BUILD files in the "peripheral" support tier

Tom Stellard via llvm-dev llvm-dev at lists.llvm.org
Tue Dec 8 19:31:19 PST 2020


On 12/8/20 7:19 PM, Mehdi AMINI wrote:
> 
> 
> On Tue, Dec 8, 2020 at 7:09 PM Tom Stellard <tstellar at redhat.com 
> <mailto:tstellar at redhat.com>> wrote:
> 
>     On 12/8/20 5:46 PM, Mehdi AMINI wrote:
>      >
>      >
>      > On Tue, Dec 8, 2020 at 12:40 PM Eric Christopher
>     <echristo at gmail.com <mailto:echristo at gmail.com>
>      > <mailto:echristo at gmail.com <mailto:echristo at gmail.com>>> wrote:
>      >
>      >     +Geoffrey Martin-Noble <mailto:gcmn at google.com
>     <mailto:gcmn at google.com>>  and +Tom Stellard
>      >     <mailto:tstellar at redhat.com <mailto:tstellar at redhat.com>>
>      >
>      >     In my effort to smooth the process out here I spoke with Tom
>     offline
>      >     and we've agreed that a pitch proposal seems to be the best way
>      >     forward.  From our discussion I believe that he disagrees with
>      >     adding unsupported build systems to llvm and what methodology we
>      >     should use to determine their or similar multiple versions of
>      >     functionality inclusion (please do correct me if I'm wrong
>     here). I
>      >     think it makes sense to limit the discussion in the pitch to
>     adding
>      >     unsupported build systems.
>      >
>      >
>      > I still have strong concerns here: "he disagrees with adding
>     unsupported
>      > build systems to llvm".
>      > It seems that you're going for a pitch which is 1) not about this
>      > proposal in particular and 2) about a point that has been explicitly
>      > discussed and written down in the "community support policy".
>      >
>      > This also somehow does not align with Tom's last email in this
>     thread:
>      >
>      >  > My understanding of the policy is that these categories of things
>      > still need to be approved in order to be added to the tree.
>      >
>      > This acknowledged the policy, and I believe acknowledged that each
>      > individual proposal is discussed on its own merits. The opposition
>      > solely based on the principle of disagreeing "with adding
>     unsupported
>      > build systems to llvm" is completely off here. Why are we writing
>     our
>      > community policy as a guideline of what is/isn't OK if we have to
>      > escalate continuously?
>      > I still strongly believe that this is not an OK strategy here.
>      >
> 
>     I don't think we should have alternative build systems in tree.  I am
>     not the only one who has expressed concerns about this.  I also don't
>     think me or anyone else should be able to unilaterally NAK a proposal.
>     This is why I am suggesting that we use the LLVM Proposal Process.
> 
>     I really don't want to spend any more time debating this on the mailing
>     list, because I don't think we are making any forward progress.  If you
>     believe my objections are not in line with the existing policy, then
>     feel free to move forward with this as an RFC.
> 
> 
> I believe the policy is *very explicit* and leaves no doubt on the topic 
> right now, it explicitly mentions alternative build systems.
> I don't understand why you didn't raise your concern with respect to the 
> policy.
> 
> I mostly object to the fact that this proposal can't be discussed on 
> it's own because you're addressing your disagreement with the policy in 
> this thread and in the context of this proposal.
> If you think that we shouldn't have any alternative build systems in 
> tree, then I believe that you should start an RFC (/proposal) to amend 
> the policy instead.
> I also believe that it is also perfectly fine to put this current RFC on 
> hold in order to address your more general objection (if you want to 
> pursue such a proposal). This seems like it would have to address the 
> existence of GN build files.
> 

I'm not going to write my own RFC/proposal, so you don't need to put 
this RFC on hold.

-Tom

> Best,
> 
> -- 
> Mehdi
> 
> 
>     In the end, I'm just giving my opinion on what I think is in the best
>     interest of the community.  It is OK with me if other people
>     disagree or
>     the community decides to go another direction.  I know not everyone has
>     the same perspective as me, and different perspectives are what make
>     communities strong.
> 
>     -Tom
> 
> 
> 
> 
> 
>      > Best,
>      >
>      > --
>      > Mehdi
>      >
>      >
>      >     My personal take on this and why I've been helping shepherd
>     this along:
>      >
>      >     I believe that we should be enabling other people to do work
>     in llvm
>      >     as long as
>      >
>      >       a) it doesn't impact maintainability of the core system
>     (open to
>      >     debate in some ways),
>      >       b) they have a history/desire to be responsible
>     maintainers, and
>      >       c) it's easy enough to remove if it becomes an issue.
>      >
>      >     and that doing this helps llvm be used more easily in other
>      >     projects; thus helping see it's inclusion in more projects, a
>     goal
>      >     of the project as a whole.
>      >
>      >     Thanks!
>      >
>      >     -eric
>      >
>      >     On Sun, Dec 6, 2020 at 7:07 AM Renato Golin
>     <rengolin at gmail.com <mailto:rengolin at gmail.com>
>      >     <mailto:rengolin at gmail.com <mailto:rengolin at gmail.com>>> wrote:
>      >
>      >         On Sun, 6 Dec 2020 at 04:38, Mehdi AMINI
>     <joker.eph at gmail.com <mailto:joker.eph at gmail.com>
>      >         <mailto:joker.eph at gmail.com
>     <mailto:joker.eph at gmail.com>>> wrote:
>      >
>      >             It isn't clear to me what makes you say that? You may not
>      >             have been involved with it and you may haven't been
>     paying
>      >             attention at the time, but it seems unfair to claim
>     that it
>      >             didn't have scrutiny or it went in without the usual
>     proper
>      >             consideration.
>      >             In particular it has been discussed on llvm-dev@ like any
>      >             other proposal, and the thread was pretty long:
>      > http://lists.llvm.org/pipermail/llvm-dev/2018-October/127342.html
>      >             ; it also went further with a lightning talk **and** a
>      >             round-table during a llvm dev meeting.
>      >
>      >
>      >         Sorry, scrutiny was the wrong word. I meant "trouble". This
>      >         proposal seems to be having a lot of trouble that GN
>     should have
>      >         had too. The biggest push back is about adding new build
>      >         systems, not Bazel versus GN versus CMake.
>      >
>      >         There seems to be a conflict here about adding a
>     secondary build
>      >         system. The first could be always thought of as an exception,
>      >         but the second looks very much like a pattern. The way I
>     see it,
>      >         from one side there's people worried about maintenance and
>      >         proliferation of code that is not directly related to the
>     LLVM
>      >         project (like build systems, editor files, etc) and from the
>      >         other side, there's people saying this has been happening
>     for a
>      >         long time.
>      >
>      >         I tried to solve that by starting the support policy, but not
>      >         with the intent to validate the inclusion of GN/Bazel,
>     just to
>      >         help the discussion move to a consensus. I regret having
>     written
>      >         GN and Bazel by name, which only now I realise they could be
>      >         used as leverage for one side of the discussion. It was
>     not my
>      >         intention, and I don't think we should ignore the issues just
>      >         because GN has been included already, either.
>      >
>      >         My support for moving this to a document (not necessarily a
>      >         proposal) is because for most of the original discussion
>     around
>      >         Bazel, throughout the discussion about the support policy and
>      >         now the retake on Bazel's inclusions, Tom's points
>     haven't been
>      >         addressed completely. There seems to be more discussion
>     around
>      >         semantics, history and precedence than the actual technical
>      >         details. I'm guilty of that, too, while trying to solve the
>      >         conflict, and I apologise if the support policy has
>     created more
>      >         confusion than it solved.
>      >
>      >         I think laying out the issues in a document and
>     discussing the
>      >         technical aspects over it would make things easier, not
>     harder.
>      >         If the support policy needs to be amended to clarify
>     that, so be
>      >         it. We need to document what happens and what we want to
>     happen,
>      >         not fix some version of the past as a golden standard for the
>      >         future.
>      >
>      >         But as I always say: whatever works. If you want to continue
>      >         discussing in this thread, by all means, do go on.
>      >
>      >         cheers,
>      >         --renato
>      >
> 



More information about the llvm-dev mailing list