Let's see what CI says. I think this will work fine in clang13 onward which is
all we care about soon.
Details
- Reviewers
philnik • Quuxplusone - Group Reviewers
Restricted Project - Commits
- rG8f0b2ac140c3: [libc++] Use _LIBCPP_NO_UNIQUE_ADDRESS for base in join_view
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
libcxx/include/__ranges/join_view.h | ||
---|---|---|
71 | Was a bug ever filed? If not, please do; if so, please give a link in the commit message. FWIW, in the absence of any other information, I'd trust the old code comment that says "don't do this" over the CI results. ;) But I'm very curious to understand what the old code comment was talking about, because how on earth could the attribute be OK on __cache_ but ICE the compiler on __base_?? |
libcxx/include/__ranges/join_view.h | ||
---|---|---|
71 | I don't understand how/why this could crash clang. I haven't seen a relevant bug report filed (nor am I convinced it is an actual bug). The closest one I could find is https://github.com/llvm/llvm-project/issues/50766. @zoecarver @ldionne - any ideas here? |
Please update to use _LIBCPP_NO_UNIQUE_ADDRESS, but then, in the absence of any further information, I think we should just land this and see what breaks. My bet is still on "nothing." :)
Was a bug ever filed? If not, please do; if so, please give a link in the commit message.
FWIW, in the absence of any other information, I'd trust the old code comment that says "don't do this" over the CI results. ;) But I'm very curious to understand what the old code comment was talking about, because how on earth could the attribute be OK on __cache_ but ICE the compiler on __base_??