I received a report that MCE is hanging for a CPAN smoke tester and possibly others [SOLVED]

[marioroy]

Hi PerlMonks,

I'm the author of Many-core Engine (MCE) for Perl.

I received an email stating that MCE is hanging for him and possibly other smoke testers out there. The unfortunate thing is that I'm not able to reproduce the issue. What should I do?

This is the message I received

Your MCE perl module's tests hang on all of my CPAN-testing machines:

PERL_DL_NONLAZY=1 /home/david/cpantesting/perl-5.8.9/bin/perl
  "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib',
  'blib/arch')" t/*.t
t/00_required_modules.t .... ok
t/00_required_signals.t .... ok
t/01_load_mce.t ............ ok
[hangs here]

It's not just that the tests are taking a long time, because if I just leave it there CPU usage and I/O are minimal. I see that there have been no CPAN-testers reports for the module at all, so I guess that it's hanging for everyone else too, and this isn't just an artifact of my environment.

End of message

The last reported (passing) is 1.008. No reports on 1.1x have appeared at http://www.cpantesters.org/distro/M/MCE.html. MCE 1.102 is seen passing at http://code.activestate.com/ppm/MCE/. MCE (any versions) never passed for Windows (32-bit) at ActiveState and I'm aware of that.

Need your guidance on what to do. I'm not able to reproduce MCE hanging on my end like the tester is seeing. And it's hanging where I'm only loading the Module. Really not sure why it would hang there like that.

Thank you...

[marioroy]

Thanks to the CPAN tester by the name David Cantrell, he allowed me to log into his system using a test account.

Issue is resolved.  I'm going to release an update.

Inside Signal.pm, added writeable check -w '/dev/shm'

   $_tmp_dir_base = ($_use_dev_shm && -d '/dev/shm' && -w '/dev/shm')
      ? '/dev/shm' : '/tmp';

Also in test script, also added writeable check -w '/dev/shm'

   use_ok('MCE::Signal', qw( $tmp_dir -use_dev_shm ));

   if (! exists $ENV{TEMP} && -d '/dev/shm' && -w '/dev/shm')
      ...

I feel so terribly bad that MCE is causing smoke testing to hang out there. I'm thankful for David Cantrell who reported on the issue to me via an email.

Do any folks have stories where something like this such as stalling smoke testing? Oh, how I feel horrific at the moment. :(

MCE 1.103 should be up on CPAN momentarily.

Mario

[davido]

It happens. I once had a 'prompt' that wasn't supplying a valid default in non-interactive environments, and that was causing a stall. It's fixed now, and I learned something in the process:

When a smoker sees your script causing their rig to hang, they add your module to a blocklist so that it doesn't get tested in the future. They're happy to remove it once you've got the problem fixed, but you will need to alert them to do so. Visit http://lists.perl.org and subscribe to [e-mail]. Then send a message requesting that smokers begin testing your module again, starting with version XYZ. Explain that good behavior has been re-established, and request removal from their blocklists.

The mailing list is low traffic, but it might not be inappropriate to follow up a week later mentioning the newfound success, and reminding any who haven't removed your module from their ban list to go ahead and do so.

[marioroy]

MCE-1.103 has been released on CPAN. Running cpan -t MCE is still pulling down 1.102. I will send a message to the list once I see MCE 1.103 being tested.

Thanks for the helpful feedback Dave.

[marioroy]

For newbies like myself on perlmonks.org, David Cantrell is known as DrHyde.

Thank you DrHyde for providing a test account. I was able to reproduce the issue.

Shame on me for making the assumption that /dev/shm is writable by everyone and the reason for MCE 1.103.

Regards,

Mario