[PATCH] D59038: [8.0 Regression] Fix handling of `__builtin_constant_p` inside the enable_if attribute.

Fri Mar 8 11:16:33 PST 2019

EricWF added a comment.

In D59038#1422905 <https://reviews.llvm.org/D59038#1422905>, @hans wrote:

> In D59038#1422890 <https://reviews.llvm.org/D59038#1422890>, @rsmith wrote:
>
> > LGTM, and I think this is safe enough to take for Clang 8.
>
>
> Do you think the severity is high enough to spin another release candidate?

Yes. Very much so. It also effect template arguments, case statements, enumerators, and probably more.

> My concern is that since this didn't show up in testing until now, which I guess is quite a while after it regressed, maybe it's not important enough to delay the already late release?

The common case for hitting this bug requires using libstdc++ 7.0 or newer and C++17. This may be relatively rare for our release testers, who are probably either using libc++, targeting the system libstdc++, or not turning on C++17.
As libstdc++ 7.0 becomes more common and as more users move to C++17 people are going to run into this a lot.

> Also it sounds like it isn't completely fixed yet..

It is now.

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D59038/new/

https://reviews.llvm.org/D59038