[PATCH] Async signal safe TLS accesses

[Carlos O'Donell]

On 11/30/2013 01:32 PM, Rich Felker wrote:
> I'm against any attempt to make anything "safe" under async
> cancellation. Calling any library functions at all (other than to
> disable async cancellation) with async cancellation active is just a
> serious programming error. There are way too many things that can go
> wrong, and no good reason to do it.

I think we should make pthread_testcancel AS-Safe and guarantee that
in glibc to fix the POSIX design flaw and then file a bug with the
Austin group to have it on record (even if its promptly closed).

There are some cases where AS-Safety is useful, and I'd like to see
that small bubble extended just a little bit.

This is however an entirely different topic form the one at hand,
but since I saw your comment about pthread_testcancel, I wanted to
put the bug in your ear.

Cheers,
Carlos.