Re: kern/58091 (after fork/execve or posix_spawn, parent kill(child, SIGTERM) has race condition making it unreliable)

[riastradh%NetBSD.org@localhost]

Synopsis: after fork/execve or posix_spawn, parent kill(child, SIGTERM) has race condition making it unreliable

State-Changed-From-To: open->analyzed
State-Changed-By: riastradh%NetBSD.org@localhost
State-Changed-When: Thu, 13 Mar 2025 01:54:36 +0000
State-Changed-Why:
This turns out to be a rather embarrassing self-inflicted bug where
execve(2) and posix_spawn(2) first carefully discard only those pending
signals whose dispositions are being reset to a default action of being
ignored (e.g., SIGCHLD, SIGWINCH, ...)...and then, in a separate
subroutine, just blithely discard _all_ pending signals anyway.  The
two subroutines were written at different times, which explains some of
the confusion, and one of them can be entirely deleted, which I will do
once the releng test bed runs the new tests I added for this.

Sorry for the delay in handling this issue -- it escaped our attention
when it first turned up, apparently, or whoever investigated when it
first arrived came up empty-handed and didn't pursue it.