Hi Yuyi,
On Sat, 8 Mar 2025 16:43:05 +0800
Yuyi Wang wrote:
>
> The current behavior returns EINVAL on these 2 signals, which is
> different from the requirement of POSIX. In addition, it makes
> posix_spawn fail to set POSIX_SPAWN_SETSIGDEF. In
> newlib/libc/posix/posix_spawn.c:200, it tries to set for all signals
> including SIGKILL & SIGSTOP, and sigaction should not fail.
> ---
> winsup/cygwin/signal.cc | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/winsup/cygwin/signal.cc b/winsup/cygwin/signal.cc
> index f8ba67e75..a964c3b29 100644
> --- a/winsup/cygwin/signal.cc
> +++ b/winsup/cygwin/signal.cc
> @@ -440,7 +440,7 @@ sigaction_worker (int sig, const struct sigaction
> *newact,
> oa.sa_handler);
> if (sig == SIGKILL || sig == SIGSTOP)
> {
> - set_errno (EINVAL);
> + res = 0;
> __leave;
> }
> struct sigaction na = *newact;
> --
> 2.48.0
>
Thanks for the patch. However, Corinna and I do not think this is
the right thing. sigaction() returns EINVAL for SIGKILL/SIGSTOP
on Linux as well.
I disccussed with Corinna, and concluded the problem is in posix_spawn()
rather than sigaction(). posix_spawn() must not try to call sigaction()
on SIGKILL/SIGSTOP.
Would you like to create a patch for posix_spawn()? Or shall we do that?
--
Takashi Yano <[email protected]>
