[libcxx-commits] [PATCH] D97742: [C++23] [P1518R2] Better CTAD behavior for containers with allocators

[Louis Dionne via Phabricator via libcxx-commits]

ldionne accepted this revision.
ldionne added a comment.
This revision is now accepted and ready to land.

I spoke to Arthur about this offline. We usually don't implement stuff before it gets into the Standard, and I'd like to avoid creating a precedent. Waiting for it to get into the standard also means that we won't be caught off-guard if the paper changes at the last minute (which happens), and it will simplify our book-keeping by not having to remember to go back and update the Status list after the paper lands. So instead, Arthur and I agreed that we could pre-review this and when it's ready, he will just sit on it until the paper is actually voted in plenary, and then he can commit it immediately.

We also normally don't implement extensions (e.g. back-porting a paper to an earlier standard), and there are good reasons for that (it leads to tech debt, bugs down the road and we sometimes even break users in subtle ways). However, in this case, I think it's really reasonable to implement in previous standards:

1. This is basically fixing a defect that we hadn't noticed in previous standards, so it's akin to a LWG issue IMO.
2. The only way this can hurt someone is if they somehow rely on the use cases enabled by the paper not working in a SFINAE context (or at least I can't think of something else). I have little sympathy for these sorts of breakages - otherwise we'd be prevented from making basically any change.
3. Fixing the issue in previous standards increases consistency and reduces code complexity (in the test suite too), so we don't lose anything here.

The change itself looks good to me, but we'll have to edit the Status page upon landing it.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D97742/new/

https://reviews.llvm.org/D97742