rudeb0y has asked for the wisdom of the Perl Monks concerning the following question:

Just a general question....no code. Is there a way to wrap a timer around a system call.? I have a command that will timeout after 5 minutes. Is there a way to make the subroutine go to the next element for processing...say after 2 minutes.?? Any help/leads would be helpful.

Replies are listed 'Best First'.
Re: Timer Question
by ikegami (Patriarch) on Feb 16, 2006 at 19:28 UTC
    fork a child, then monitor (alarm+waitpid) the child in the parent. I'm no expert at working with child processes, but I came up with the following. Once the child has run for $timeout seconds, the parent asks the child to terminate. If the child is still running 10 seconds later, the parent forcefully kills the child.
    my @command = "sleep 10; echo awoke"; my $timeout = 5; my $cpid = fork(); die("Unable to fork: $!\n") if not defined $cpid; if (not $cpid) { # In child exec(@command); die("Unable to exec: $!\n"); } eval { local $SIG{ALRM} = sub { die "alarm\n" }; alarm($timeout); waitpid($cpid, 0); }; alarm(0); if ($@) { die $@ if $@ ne "alarm\n"; warn("Long running child. Attempting to terminate.\n"); kill 'TERM', $cpid; eval { local $SIG{ALRM} = sub { die "alarm\n" }; alarm(10); waitpid($cpid, 0); }; alarm(0); if ($@) { die $@ if $@ ne "alarm\n"; warn("Long running child. Attempting to kill.\n"); kill 'KILL', $cpid; eval { local $SIG{ALRM} = sub { die "alarm\n" }; alarm(5); waitpid($cpid, 0); }; alarm(0); if ($@) { die $@ if $@ ne "alarm\n"; warn("Unable to kill long running child.\n"); } else { warn("Long running child successfully killed.\n"); } } else { warn("Long running child successfully terminated.\n"); } }
    Output:
    $ perl script.pl Long running child. Attempting to terminate. Long running child successfully terminated.
Re: Timer Question
by saintmike (Vicar) on Feb 16, 2006 at 18:47 UTC

      That's didn't work prefectly.

      $ perl -e 'alarm(5); system("sleep 10; echo awoke")' Alarm clock $ awoke

      system was interupted, but the command kept executing.

        Depends on what is meant by "system call".

        If it's system call, it's a call to the OS, like sysread, in which case alarm works exactly as required.

        If it's "system() call", then things are different: system() forks a process, executes whatever's specified (different for single/multiple arguments), waits until the command is completed and then returns. It won't kill the spawned process, that's why it will run until the end.