[Bug libc/22273] Improper assert in Linux posix_spawn implementation

["fweimer at redhat dot com" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=22273

--- Comment #4 from Florian Weimer <fweimer at redhat dot com> ---
(In reply to Adhemerval Zanella from comment #3)

> However I do not think setting the err atomically would help here, ideally
> we will need kernel help on to signaling abrupt child execution. One option
> would be to set all signals to SIG_IGN (although it won't help with SIGKILL
> or SIGSTOP).

I think atomic accesses are needed for their compiler barrier properties.

> What about removing the assert and setting all signal handler as SIG_IGN?

We can't do that because the signal handler disposition is inherited across
execve.  I think it is sufficient to check the waitpid status in combination
with the error code.  I don't think we need to report whether the subprocess
was terminated before or after the execve (which can be successful or not).

-- 
You are receiving this mail because:
You are on the CC list for the bug.