MaskRay added a comment. In D151683#4409633 <https://reviews.llvm.org/D151683#4409633>, @aaron.ballman wrote:
> In D151683#4384017 <https://reviews.llvm.org/D151683#4384017>, @erichkeane > wrote: > >> In D151683#4382408 <https://reviews.llvm.org/D151683#4382408>, @philnik >> wrote: >> >>> No. I guess, since you are asking I should write one for this? Only for the >>> removal of `-fdouble-square-bracket-attributes`, or also for back porting >>> this? >> >> The RFC I would expect for "allow C/C++ style attributes as an extension in >> previous modes". This is a pretty significant feature to allow as an >> extension, particularly since our two main alternative compilers (G++/MSVC) >> don't have this as an extension. I'd be curious how the other C compilers do >> this as well. > > I think this does warrant an RFC because it impacts both C++ and C, but I'm > hoping the RFC is uncontroversial. It's mostly to establish what the use > cases are for needing the extension. The feature is conforming as an > extension to both languages, and I don't know of any breakage that would come > from enabling it by default. I'm not certain whether we want to remove the > feature flag immediately or whether we'd like to give one release of warning > about it being removed (I'll defer to whatever @MaskRay thinks is reasonable) > -- that search is compelling for why it's safe to remove the option, but > there's plenty of proprietary code which we don't know about, so it's > possible the option is still used. I'd be especially curious to know if > anyone is using `-fno-double-square-bracket-attributes` to disable the > feature in a mode where it would otherwise be enabled. I'm adding the > `clang-vendors` group as a subscriber as an early warning that removing the > command line option could be a potentially breaking change. > > In terms of implementation work, there's still some minor stuff to address. > Also, please be sure to also add a release note about the changes, and a > potentially breaking entry for removing the command line option (assuming we > elect to go that route). If allowing the extension in older language modes deems good, removing `-fno-double-square-bracket-attributes` seems fine if (and thanks for CCing the `clang-vendors` group). If we want to play safely, we can default `-fdouble-square-bracket-attributes` to true in this patch and remove the option in a follow-up change possibly after some time. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D151683/new/ https://reviews.llvm.org/D151683 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
