Yes there is, it's named alarm. I'm not exactly sure what you mean by a remote shell... are you using Expect, or Net::Telnet or system('sh') or something else entirely? Nevertheless, your code is going to look something like this:

eval {
  local $SIG{ALRM} = sub { die "too long\n" };
  alarm 30; #seconds
  system( '/usr/local/bin/scp', 'me@example.com:foo.txt', '.' );
  alarm 0;
};
if ($@ eq "too long\n") {
  # time out
}
elsif ($@) {
  #something else went wrong
}
else {
  # ok
}
[download]

Note that your remote command could involve a Win32 machine, however you cannot run this code on a Win32 machine as alarm is not implemented on that platform. Also not that you want your signal handler $SIG{ALRM} to be as simple as possible. The more it does, the more chance you run the risk of dumping core and other other unpleasant things

you cannot run this code on a Win32 machine as alarm is not implemented on that platform.

instead, use Win32::Event. there's an example at ALRMing behavior on Win32

~Particle *accelerates*

--t. alex

"Mud, mud, glorious mud. Nothing quite like it for cooling the blood!" --Michael Flanders and Donald Swann

Both of these examples are fine for stuff that doesn't do any calls to system or exec.

But that's probably exactly what he wanted to do.
The quick answer is: Read perlipc and go on from there.

Long answer: working on that ;-)

Update: Here's the long answer:

#!/usr/bin/env perl

use strict;
use warnings;
use POSIX qw/:sys_wait_h/;

my $child_pid = undef;

my $status = fork;
die "failed to fork\n" unless defined $status;

if ($status > 0) {
  # this is the parent process.
  local $SIG{ALRM} = sub { print STDERR "operation timed out for PID $
+status\n"; kill 9, $status };
  alarm 10;
  my $pid = 0;
  do {
    $pid = POSIX::waitpid ($status, POSIX::WNOHANG);
  } until ($pid > 0);
  alarm 0;
} else {
  # this is the child process.
  system "ssh a_server_could_hang";
}
[download]

For any non-portability hit your OS vendor :-(

Kay

Your short answer is correct, but your solution doesn't work on AIX at least. It suffers from the same problem as the normal 'alarm' solution -- the parent dies, but the child (or grandchild in this case) lives on.

The kill doesn't kill the grandchild created by system (even if you use -9 instead of a 9). Also, I'm not sure why you are using an alarm loop around a non-blocking waitpid. It's just going to spin the CPU for 10 seconds. You should either use alarm with a blocking waitpid or just poll/sleep/poll with the non-blocking one.

bluto