Re: [PATCH 4/5] linux: Optimize posix_spawn spurious sigaction calls

[Florian Weimer <fweimer at redhat dot com>]

* Adhemerval Zanella:

> But still I don't see that this being more complex to backport as being
> a impeding reason to push it upstream.

Your argument was that the glibc-only optimization would be available to
users more quickly.  That's why I brought backporting up.

I still think the direction this optimization is taking quite wrong.  We
shouldn't add code to system call wrappers to collect secondary
information if we can help it.  openat had something like this, and it
went wrong with O_TMPFILE.  sigaction is not really simple, either.  It
has its own flags, and the kernel might enhance the system call in
unexpected ways, too.

Just to be clear, I think it's worthwhile to optimize this for the
reasons you indicated.  We merely disagree about the means. 8-)

Thanks,
Florian