file read timeout

[0xbeef]

The following code snippet provides a crude timer mechanism for file read failures, somewhat along the lines of what Re: Redirecting stdout/stderr to pipe achieves for commands. The real code will run on many systems and I deem the timeout as necessary, although I know read failures are uncommon.

I eliminated the use of sysread+select due to the fact that I need @$outref populated line-by-line (whereas sysread uses buffers). I also do not have any real ideas on encoding. Can anyone make any improvement suggestions or is this ok?

Niel

UPDATE: fixed an endless loop in the example.

#!/usr/bin/perl -w
use strict;
use IO::Select;
use Symbol qw(gensym);

my $g_verbose = 1;
my $g_pid;
my $g_stop;
my $g_timeout = 2; #for test purposes
my $g_did_run;
my $g_max_cmd_size = 2_000_000;
my $g_max_file_size = 2_000_000;
my %g_pids;

sub read_stat($$$) {
   my ($file,$rcref,$statref) = @_;
   @$statref = stat($file);
   if ( $#$statref < 0 ) {
      print "Stat of file '$file' failed: $!\n";
      $$rcref = 1;
   } else {
      $$rcref = 0;
   }
   return $$rcref;
}
sub read_file($$$$$) {
   my ($file,$encoding,$rcref,$outref,$statref) = @_;
   @$outref = ();
   my $outbytes = 0; 
   $g_stop = 0;

# die if timeout occurs!
   local $SIG{ALRM} = sub {
      print "File \'$file\' collection reached deadline after $g_timeo
+ut secs.\n";
      die "Exiting - a read operation timed out!\n";
      $g_stop = 1;
   };
   alarm($g_timeout);

   if ( !defined($encoding) ) {
      $encoding = 'text/plain';
   }
   if ( $encoding ne 'text/plain' ) {
      print "Error: Can't collect file '$file' - '$encoding' encoding 
+is not supported.\n";
      $$rcref = -2;
      return $$rcref;
   }
   if ( $g_verbose ) {
      print "Collecting '$file'.\n";
   }

   read_stat($file,$rcref,$statref);
   if ( $$rcref != 0 ) {
      return undef;
   }
   if ( !open(IFILE,"<$file") ) {
      print "Open of file '$file' failed: $!\n";
      return undef;
   }

# read data into @$outref (line by line). alarm prevents timeout.
   while (!$g_stop) {
      last if (eof(IFILE));
      my $line = <IFILE>;
      chomp $line;
      push @$outref, $line;
# for test purposes - make it timeout!
      sleep 5;
      $outbytes += length($line);
      if ($outbytes >= $g_max_file_size) {
         print "Maximum file size reached after $outbytes bytes\n";
         last;
   alarm(0);
   close(IFILE);

   if (!$g_stop) {
      $$rcref = 0;
   } else {
      $$rcref = -1;
   }
   return $$rcref;
}
#MAIN
my $file = '/etc/hosts';
my $encoding = 'text/plain';
my $rc;
my (@cmdout,@stat);
$rc = read_file($file,$encoding,\$rc,\@cmdout,\@stat);
print "Collected $file (rc=$rc)\n";

[cdarke]

You seem to be missing at least two close curly brackets, one after the while loop and one after the if statement before the alarm.

You return undef on error, so the final statement might be better written as:
print "Collected $file (rc=". (defined $rc?$rc:'undef') . ")\n";
You are mixing stdout (bare print) and stderr (die) messages. I suggest you only print error messages to stderr, use print STDERR, or warn, or incorporate into the die statement that follows.

Your while loop could be simplified in a number of ways. For example:


while (my $line = <FILE>) {
         chomp $line;
         push @$outref, $line;
   # for test purposes - make it timeout!
         sleep 5;
         $outbytes += length($line);
         if ($outbytes >= $g_max_file_size) {
            warn "Maximum file size reached after $outbytes bytes";
            last;
         }
   }


Assuming I put the close braces in the right place.
Another approach might be to avoid the loop altogether by testing the file size before you start reading:


   my $file_size = -s IFILE;
   die "$file too large at $file_size bytes" if $file_size > $g_max_fi
+le_size;
  @$outref = <IFILE>;