Async-signal-safe access to __thread variables from dlopen()ed libraries?

[Paul Pluzhnikov]

On Wed, Jun 13, 2012 at 11:18 AM, Richard Henderson <rth@twiddle.net> wrote:

> I don't see why a linker change is necessary.  The only reason we do so
> now is to enable linker relaxation.  It doesn't seem like the situations
> in which this new builtin would be useful would be conducive to relaxation.

Yes, I think you are right.

-- 
Paul Pluzhnikov