bin/59289: atf should report more detailed signal info on crash

[campbell+netbsd%mumble.net@localhost]

>Number:         59289
>Category:       bin
>Synopsis:       atf should report more detailed signal info on crash
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    bin-bug-people
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Sun Apr 13 15:00:00 +0000 2025
>Originator:     Taylor R Campbell
>Release:        current, 10, 9, ...
>Organization:
The NetATF Foundasignal
>Environment:
>Description:
When a test program terminates on a signal, atf just prints the signal number, e.g.:

Test case: lib/libc/regex/t_exhaust/regcomp_too_big

Duration: 26.751680 seconds
Termination reason

FAILED: Test program received signal 9

https://releng.netbsd.org/b5reports/evbarm-aarch64/2025/2025.04.11.23.57.21/test.html#lib_libc_regex_t_exhaust_regcomp_too_big

It would be nice if it additionally printed:

- the identifier of the signal (SIGKILL, in this case, for those of us who can't remember the numbers -- OK, I remember 9 is SIGKILL, but not the other ones)
- the short description of the signal (SIGKILL is `Killed'; SIGFPE is `Floating-point exception'; &c.)
- relevant information from the siginfo_t
>How-To-Repeat:
browse atf reports with terse signal termination information
>Fix:
Yes, please!

The identifier is available (minus the `SIG' prefix) via signalname(3) in NetBSD, or via sig2str in POSIX.1-2024 (not yet implemented in NetBSD):

https://man.NetBSD.org/signalname.3
https://pubs.opengroup.org/onlinepubs/9799919799/functions/sig2str.html

The short description is available via the array sys_siglist of length sys_nsiglist, indexed by the signal number, in NetBSD.  Not sure if there's any portable analogue of this.

https://man.netbsd.org/sys_siglist.3

The siginfo_t is more elaborate and documentd in the siginfo(3) man page:

https://man.netbsd.org/siginfo.2

This is helpful for quickly identifying types of crashes, like alignment errors vs hardware mapping errors, unmapped access vs forbidden access, &c., and at what address the error happened.