Why does my jzip process hang when I call it with Perl's system?

Question

I am definitely new to Perl, and please forgive me if this seem like a stupid question to you.

I am trying to unzip a bunch of .cab file with jzip in Perl (ActivePerl, jzip, Windows XP):

#!/usr/bin/perl

use strict;
use warnings;

use File::Find;
use IO::File;

use v5.10;

my $prefix = 'myfileprefix';
my $dir = '.';

File::Find::find(
    sub {
        my $file = $_;
        return if -d $file;          
        return if $file !~ /^$prefix(.*)\.cab$/;  
  my $cmd = 'jzip -eo '.$file;
  system($cmd);
    }, $dir
);

The code decompresses the first .cab files in the folder and hangs (without any errors). It hangs in there until I press Ctrl+c to stop. Anyone know what the problem is?

EDIT: I used processxp to inspect the processes, and I found that there are correct number of jzip processes fired up (per the number of cab files resides at the source folder). However, only one of them is run under cmd.exe => perl, and none of these process gets shut down after fired. Seems to me I need to shut down the process and execute it one by one, which I have no clue how to do so in perl. Any pointers?

EDIT: I also tried replacing jzip with notepad, it turns out it opens up notepad with one file at a time (in sequential order), and only if I manually close notepad then another instance is fired. Is this common behavior in ActivePerl?

EDIT: I finally solved it, and I am still not entire sure why. What I did was removing XML library in the script, which should not relevant. Sorry I removed "use XML::DOM" purposefully in the beginning as I thought it is completely irrelevant to this problem. OLD: use strict; use warnings;

use File::Find;
use IO::File;
use File::Copy;
use XML::DOM; 

use DBI;
use v5.10;

NEW:

#!/usr/bin/perl
use strict;
use warnings;

use File::Find;
use IO::File;
use File::Copy;

use DBI;
use v5.10;

my $prefix = 'myfileprefix';
my $dir = '.';

# retrieve xml file within given folder
File::Find::find(
    sub {
        my $file = $_;
        return if -d $file;             
        return if $file !~ /^$prefix(.*)\.cab$/;
        say $file;
        #say $file or die $!;
        my $cmd = 'jzip -eo '.$file;
        say $cmd;
        system($cmd);       
    }, $dir
);

This, however, imposes another problem, when the extracted file already exists, the script will hang again. I highly suspect this is a problem of jzip and an alternative of solving the problem is simply replacing jzip with extract, like @ghostdog74 pointed out below.

Answer 1

First off, if you are using commands via system() call, you should always redirect their output/error to a log or at least process within your program.

In this particular case, if you do that, you'd have a log of what every single command is doing and will see if/when any of them are stuck.

Second, just a general tip, it's a good idea to always use native Perl libraries - in this case, it may be impossible of course (I'm not that experienced with Windows Perl so no clue if there's a jzip module in Perl, but search CPAN).

UPDATE: Didn't find a Perl native CAB extractor, but found a jzip replacement that might work better - worth a try. http://www.cabextract.org.uk/ - there's a DOS version which will hopefully work on Windows

Answer 2

Based on your edit, this is what I suggest:

#!/usr/bin/perl

use strict;
use warnings;

use File::Find;
use IO::File;

use v5.10;

my $prefix = 'myfileprefix';
my $dir = '.';

my @commands;

File::Find::find(
    sub {
        my $file = $_;
        return if -d $file;             
        return if $file !~ /^$prefix(.*)\.cab$/;        
        my $cmd = "jzip -eo $File::Find::name";
        push @commands, $cmd;
    }, $dir
);


#asynchronously kick off jzips
my $fresult;
for @commands
 {
    $fresult = fork();
    if($fresult == 0) #child
    {
     `$_`;
    }
    elsif(! defined($fresult))
    {
       die("Fork failed");
    }
    else
    {
       #no-op, just keep moving
    }

 }

edit: added asynch. edit2: fixed scope issue.

Answer 3

What happens when you run the jzip command from the dos window? Does it work correctly? What happens if you add an end of line character (\n) to the command in the script? Does this prevent the hang?

Answer 4

here's an alternative, using extract.exe which you can download here or here

use File::Find;
use IO::File;    
use v5.10;    
my $prefix = 'myfileprefix';
my $dir = '.';    
File::Find::find({wanted => \&wanted}, '.');
exit;
sub wanted {
    my $destination = q(c:\test\temp);
    if ( -f $_ && $_=~/^$prefix(.*)\.cab$/ ) {
         $filename = "$File::Find::name";
         $path = "$File::Find::dir";
         $cmd = "extract /Y $path\\$filename /E /L $destination";
         print $cmd."\n";
         system($cmd);
    }
} $dir;

Answer 5

Although no one has mentioned it explicitly, system blocks until the process finishes. The real problem, as people have noted, is figuring out why the process doesn't exit. Forking or any other parallelization won't help because you'll be left with a lot of hung processes.

Until you can figure out the issue, start small. Make the smallest Perl script that demonstrates the problem:

#!perl
system( '/path/to/jzip',  '-eo', 'literal_file_name' ); # full path, list syntax!
print "I finished!\n";

Now the trick is to figure out why it hangs, and sometimes that means different solutions for different external programs. Sometimes you need to close STDIN before you run the external process or it sits there waiting for it to close, sometimes you do some other thing.

Instead of system, you might also try things such as IPC::System::Simple, which handles a lot of platform-specific details for you, or modules like IPC::Run or IPC::Open3.

Sometimes it just sucks, and this situation is one of those times.