I'm doing repetitive things, some of which will exceed a timeout. I put the problem children in an eval block with an alarm handler to try to catch the timeouts.

This seems to work for the first hang, and continues to process 'good' entries, but the second (and subsequent) hangs don't trigger the alarm.

Any help much appreciated!

Here's a simplified version:

#!/usr/bin/perl -w
use strict;

foreach my $loop ( 1,1,6,1,6,1,6,6 ) {
  eval {
    local $SIG{ALRM} = sub { die("TIMEOUT\n") };
    alarm(5);
    my $pid = open(READER, "ping -s localhost 64 $loop |") or die("Can
+'t ping: $!\n");
    while (<READER>) {
      print;
    }
    close READER;
    alarm(0);
  };

  if ( $@ =~ /TIMEOUT/ ) {
    print "***\n";
    print "*** Caught alarm from eval block.\n";
    print "***\n";
  }
}
[download]

I'm probably doing horrible things including but not limited to mishandling the close()?

Update How very odd. I just ran this under 5.8.6 on my Mac, and it did exactly what I expected it to. (I had to change the ping options but hey). Despite the typos, it still dies on Solaris under 5.6.1?

Update 2 This kicked off some deja vu, and I looked at my past postings. Could this be related to this node?