On Tue, Jul 26, 2016 at 2:25 AM, Matt Arsenault via llvm-dev < llvm-dev at lists.llvm.org> wrote:> +1, some sort of defined process is definitely needed. In particular I > think there should be a guideline for when the backend has been reviewed > "enough" before it can be committed. >IMHO, this is a very important point missed from the discussion so far. What if a new back-end ticks all the right boxes but simply not being reviewed by established members of the community? I expect this to be quite common, especially with LLVM broaden in popularity and reaching beyond its traditional developer / user base. We all have busy lives and it's understandable that existing core developers / maintainers have no time nor inclination to thoroughly review every new back-end. When enough is enough and it's OK to press "commit" button? Yours, Andrey -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160729/e7708c65/attachment.html>
On 29 July 2016 at 12:09, Andrey Bokhanko <andreybokhanko at gmail.com> wrote:> What if a new back-end ticks all the right boxes but simply not being > reviewed by established members of the community?This case is clear. The code only goes in after proper review. Interests vary from case to case, and normally, pinging on the mailing list and on IRC is the only way to go. Waiting forever doesn't work, and asking a friend maintainer to "push it for you" or just pushing yourself is very much *never* appropriate. This is one of the few clear cases I can think of that would warrant a permanent commit ban for all involved. Upstream takes time, we all have to live with it. But the alternative is worse. :) cheers, --renato
Renato, I'm not speaking on ways to circumvent the review system, but rather on when enough is enough. Do you expect whole back-ends to receive the same level of code review as individual incremental patches to existing ones? If yes, by who? In case of existing back-ends, there are maintainers who understand the code and able to do a proper code review; who can do the same for new back-ends? Can we put some formal expectations into your guidelines? Yours, Andrey On Fri, Jul 29, 2016 at 3:38 PM, Renato Golin <renato.golin at linaro.org> wrote:> On 29 July 2016 at 12:09, Andrey Bokhanko <andreybokhanko at gmail.com> > wrote: > > What if a new back-end ticks all the right boxes but simply not being > > reviewed by established members of the community? > > This case is clear. The code only goes in after proper review. > > Interests vary from case to case, and normally, pinging on the mailing > list and on IRC is the only way to go. Waiting forever doesn't work, > and asking a friend maintainer to "push it for you" or just pushing > yourself is very much *never* appropriate. > > This is one of the few clear cases I can think of that would warrant a > permanent commit ban for all involved. > > Upstream takes time, we all have to live with it. But the alternative > is worse. :) > > cheers, > --renato >-------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20160729/d69ffd5b/attachment.html>