From: Tom Lane <tgl@sss.pgh.pa.us>
To: pgsql-hackers@lists.postgresql.org
Subject: Why aren't we using strsignal(3) ?
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <25757.1544983503.1@sss.pgh.pa.us>
Date: Sun, 16 Dec 2018 13:05:03 -0500
Message-ID: <25758.1544983503@sss.pgh.pa.us>
Precedence: bulk

While poking at the signal-reporting bug just pointed out by
Erik Rijkers, I couldn't help noticing how many places we have
that are doing some equivalent of this ugly dance:

#if defined(HAVE_DECL_SYS_SIGLIST) && HAVE_DECL_SYS_SIGLIST
    {
        char        str2[256];

        snprintf(str2, sizeof(str2), "%d: %s", WTERMSIG(exitstatus),
                 WTERMSIG(exitstatus) < NSIG ?
                 sys_siglist[WTERMSIG(exitstatus)] : "(unknown)");
        snprintf(str, sizeof(str),
                 _("child process was terminated by signal %s"), str2);
    }
#else
        snprintf(str, sizeof(str),
                 _("child process was terminated by signal %d"),
                 WTERMSIG(exitstatus));
#endif

(Plus, there's at least one place that *should* be doing this and is not.)

Not only is this repetitive and unreadable, but it's also obsolete:
at least as far back as POSIX:2008, there's a function strsignal()
that you're supposed to use instead.

I propose to replace all these places with code like

        snprintf(str, sizeof(str),
                 _("child process was terminated by signal %d: %s"),
                 WTERMSIG(exitstatus), pg_strsignal(WTERMSIG(exitstatus)));

where pg_strsignal is a trivial wrapper around strsignal() if that
exists, else it uses sys_siglist[] if that exists, else it just
returns "unrecognized signal".

			regards, tom lane