Re: How to make sure that signal handlers are run?
Roland McGrath wrote:
>
> SA_NODEFER and SA_RESETHAND are the ones that I called boneheaded,
> and the ones that are trivial to implement. That's all I was referring to.
>
> SA_NOCLDSTOP is defined--that is the only specified in 1003.1-1990.
> I don't know how you overlooked it.
Yes, it is there. I remember I saw it, I just forgot to paste it into
the message.
> SA_ONSTACK is defined but not actually implemented. (It can't be done
> complete robustly without some kernel help.)
>
> SA_NOCLDWAIT is a recent invention by somebody or other. We don't define
> it. It could be implemented without much trouble, by adding proc support
> for it (see how SA_NOCLDSTOP is implemented). But Linux 2.4.3 doesn't
> implement it, so clearly noone cares about this one.
>
> That leaves SA_SIGINFO, which is an invention of 1003.1-1996. This one is
> actually useful, and it is nontrivial. I don't expect this to be
> implemented until the signal code gets overhauled along with a pthreads
> implementation.
Ok, since I'm at least pretending to be working on pthreads and reading
the signal code, I'd like to hear your ideas on what needs to be changed.
Igor
Reply to: