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

[Carlos O'Donell]

On 09/20/2013 05:19 PM, Ian Lance Taylor wrote:
>> All I can say is that if you stick around we'll see it through to
>> some kind of solution.
> 
> Thanks, I'm glad to hear it.  That is not the message I took away from
> your earlier e-mail messages, which to me sounded like a set of
> reasons not to fix this problem.

If that's not the message you took away then I apologize for being unclear.

My previous email was intended to be a list of all the things I think we
need to consider and answer to end up with a high quality implementation.

As part of the technical discussion you can feel free to disagree or
disregard all or some of the points on my list.

> Paul said: we have a patch and asked "Is this something that has a
> chance of being acceptable into trunk?"  I expected the answer to that
> question to be "Yes, there is a good chance."  I did not expect the
> answer to be "Being lazy and dumb I don't want to maintain a complex
> TLS implementation, and being dumb means I can't."

That was a tongue and cheek response to your point about being lazy.

I'm trying to be as honest as possible with you, and I felt it
dishonest to say "Yes, there is a good chance." without ever having
seen the code for the implementation.

Are we interested in fixing things? Yes.

Would we like to fix TLS use in signals? Yes.

Is there a chance that it gets accepted into trunk? It depends.

Next step: Post code. Review. Iterate.

Cheers,
Carlos.