Re: factor/factor-parallel test fails due to 'split' changes
From:
Bruno Haible
Subject:
Re: factor/factor-parallel test fails due to 'split' changes
Date:
Wed, 29 Oct 2025 11:25:01 +0100
Collin Funk wrote:
> The actual issue was that Gnulib's posix_spawn calls sigismember in the
> child process with NSIG. On OpenBSD and Solaris this causes it to return
> -1 and set errno to EINVAL.
>
> I pushed the attatched patch to Gnulib
Thanks! Indeed, our lib/signal.in.h says that NSIG is the
"Maximum signal number + 1." The bug was present in gnulib/lib/spawni.c
since 2008, when I borrowed the code from glibc/sysdeps/posix/spawni.c.
Meanwhile, this bug is no longer present in glibc: The new code in
glibc/sysdeps/mach/hurd/spawni.c:348: for (i = 1; i < NSIG; ++i)
glibc/sysdeps/unix/sysv/linux/spawni.c:119: for (int sig = 1; sig < _NSIG;
++sig)
is correct.
Bruno