[PATCH] D19271: Change CALL_ONCE_INITIALIZATION implementation to use static initialization

[Chris Bieneman via llvm-commits]

I’m completely in favor of re-landing the patch. I spent a lot of time working on it with Reid and David Majnemer. Having a call_once mechanism is a good way to avoid certain types of static initialization, so it would be great to have this support in LLVM.

At the time I didn’t have the ability to reproduce the issue that was reported (still probably don’t), and I was pulled off onto another project shortly after.

-Chris


> On Apr 25, 2016, at 1:00 PM, Kostya Serebryany <kcc at google.com> wrote:
> 
> kcc added a comment.
> 
> In http://reviews.llvm.org/D19271#410997, @joker.eph wrote:
> 
>> I'm tempted to commit it just to get a correct bug report...
> 
> 
> IMHO this is a good idea.
> If one of the toolchains we depend on does not support proper C++ we need to have it documented (and then, fixed).
> 
> 
> http://reviews.llvm.org/D19271
> 
> 
>