erichkeane added a subscriber: rsmith. erichkeane added a comment. In D143891#4129214 <https://reviews.llvm.org/D143891#4129214>, @hubert.reinterpretcast wrote:
> In D143891#4122660 <https://reviews.llvm.org/D143891#4122660>, @aaron.ballman > wrote: > >> This is an ABI breaking change, isn't it? (The type trait now returns >> something different than it did before, which could change instantiations or >> object layout.) > > In my opinion, this is an argument for not fixing in a fix release of a > specific version of Clang--not an argument for additional option control in > new versions. > Instantiations change all the time from changes to constant expression > evaluation, overload resolution, partial ordering rules for templates, etc. > As for object layout, I believe the language and the ABI rules for triviality > diverged quite some time ago. > > I know that the evaluation for this specific case was that we don't need to > apply ABI versioning control for this because it only affects code using > Concepts, but I think we will eventually need to (re)discover when ABI > versioning control is the appropriate tool. > > I propose an action item: Someone needed to review past application of the > ABI versioning options to see if we can extract and document workable > criteria (perhaps in https://clang.llvm.org/docs/InternalsManual.html). Before doing that amount of work, I'd see if we could convince @rsmith to just share what his rule for this was, I think it was quite reasonable the few times he enforced it, and he can likely better state it at least. Even if he were to share here, I'm sure one of us would be willing to clean up his statement and put it in the Internals Manual. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D143891/new/ https://reviews.llvm.org/D143891 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
