Unruly children...

[aaron.m]

I have a script with the following:

  unlink("/ngvob/ui_rel/.emodel_is_running.flag");
  $SIG{'CHLD'} = 'IGNORE';

This is on - don't laugh - Solaris 2.6 SPARC, running Perl 5.004_04. I can't change that; it's a legacy app that we still support.

This block is called when we're bringing down the entire system; the function is called Crash(), so we expect that things aren't working quite right, but not quite as bad as what we're seeing.

On certain occasions, the Perl script in charge of things craps the bed. The most recent example produced these messages in the log files:

  Can't exec "/ngvob/ui_rel/.emodel_is_running.flag": Interrupted syst
+em call at /path/to/myscript line 596
  SIGCHLD handler "IGNORE" not defined.

Line 596 corresponds to the "unlink" line.

I don't really understand either of these messages. "Can't exec" looks like it's trying to execute the .emodel_is_running.flag file, instead of unlink() it. And how 'IGNORE' is not defined is puzzling.

Has anyone seen behavior like this, before?

[snoopy]

Just a punt. Maybe a parent process, system shutdown or something else is sending signals that aren't being handled and are causing general havoc.

One thing to try, if you're not already, is to trap and report on incoming signals, just to see what's going on.

foreach (qw/HUP QUIT KILL TERM/) {
    my $msg = "slugged by a $_";
    $SIG{$_} = sub {warn $msg};
}