Beefy Boxes and Bandwidth Generously Provided by pair Networks
Keep It Simple, Stupid

Getopt::Euclid victim of 5.10 upgrade

by scott\b (Novice)
on Feb 13, 2009 at 00:33 UTC ( #743501=perlquestion: print w/replies, xml ) Need Help??
scott\b has asked for the wisdom of the Perl Monks concerning the following question:

My problem is I use Getopt::Euclid extensively and after I upgraded to perl 5.10 on cygwin it broke. Cracking open the code it looks like the extended regex construct (?{}) is the source of my woes. In the perlre docs that construct has the following notice:

"WARNING: This extended regular expression feature is considered experimental, and may be changed without notice. Code executed that has side effects may not perform identically from version to version due to the effect of future optimisations in the regex engine."

I guess I was warned, but I had no idea Getopt::Euclid was built on such a fragile construct.

All my options look bad, so I hope someone sees an alternative to the following:

  1. Downgrade to perl 5.8 -- sucks because I'd have to give up // operator, among other things.
  2. Rewrite Getopt::Euclid's regex code -- Danger, danger Will Robinson.
  3. Switch to another Getopt module -- painful refactoring.
  4. Build custom version of perl -- possible vain hope that compiler optimization settings will affect regex actions.
  5. Switch to ActiveState perl -- best of the lot, but will lose shebang (i.e. #!) ability in cygwin scripts.

Like I said, I only see bad options. Any help or advice would be appreciated.

Scott B.

P.S. Here is the reduced test case:

use strict; use warnings; use Data::Dumper; use Config; use re 'eval'; my %ARGV; my $string = q{It is --now 12:34}; # Mimic Getopt::Euclid's generated regex to process a --now comman +d line # option. $string =~ s{(??{exists $ARGV{WHEN} }) --now\s+(\d+:\d+)(?{$ARGV{WHEN} = $^N}) }{thyme}xsm; print "Perl $Config{version} on $Config{osname}:\n", Data::Dumper->Dump( [ \%ARGV, $string ], [qw(ARGV string)] ); exit 0;

And here are the results. Note that $ARGV{WHEN} is undefined in the second (cygwin) example. That wreaks havoc on Getopt::Euclid.

: 2009-02-10 19:56:05 (C:/Perl510/bin/perl Perl 5.10.0 on MSWin32: $ARGV = { 'WHEN' => '12:34' }; $string = 'It is thyme'; : 2009-02-10 19:56:05 (perl Perl 5.10.0 on cygwin: $ARGV = { 'WHEN' => undef }; $string = 'It is thyme';

Replies are listed 'Best First'.
Re: Getopt::Euclid victim of 5.10 upgrade
by andreas1234567 (Vicar) on Feb 13, 2009 at 06:23 UTC
    Getopt::Euclid has 21 active bugs and the most recent change to the resolved bugs were 2 years ago. You depend on code that is most likely unmaintained (the assumption is mine and mine alone).

    I would look for replacements for Getopt::Euclid if I were you, even if it requires a major rewrite of your application.

    No matter how great and destructive your problems may seem now, remember, you've probably only seen the tip of them. [1]
      He could always seek the ACLs to be a maintainer himself.

      In Bob We Trust, All Others Bring Data.

      It was active when I started using it. That's part of the problem; all my new scripts over the last couple of years now use it. Unfortunately, I suspect replacing Euclid is what I'm going to have to do.
        You're an experienced user of the module, and the way you presented the problem shows you definitely have some skill. Perhaps fixing Getopt::Euclid really would be a better use of your time than reworking your application. Even if you can't/shouldn't/won't take over maintaining G::E on CPAN, you can still post your solution as a patch on rt.

        cat >~/.sig </dev/interesting

      Is there any other package that have the same function as Getopt::Euclid, i.e. parsing command line options based on the script's own POD?

      S P Arif Sahari Wibowo
        Is there any other package that have the same function as Getopt::Euclid

        I don't know. What's the problem with Getopt::Euclid ?
        Version v0.2.3 seems to be working fine for me on Cygwin 1.7.5 (perl 5.10.1).

        Getopt::Euclid is now apparently being maintained - if you have an issue with it, a bug report might be the quickest way to get that issue sorted.

Re: Getopt::Euclid victim of 5.10 upgrade
by ikegami (Pope) on Feb 13, 2009 at 01:16 UTC
    Does it work if you change "\x0E" to "^N"? Shouldn't matter, though.
      You are right. It was supposed to be (and now is) $^N. It got mangled when I cut-n-pasted the code and I did not notice it in the post preview.
        Did it help? Perl should accept both, IIRC
Re: Getopt::Euclid victim of 5.10 upgrade
by JavaFan (Canon) on Feb 13, 2009 at 10:45 UTC
    I get:
    Perl 5.8.9 on linux: $ARGV = { 'WHEN' => '12:34' }; $string = 'It is thyme';
    Perl 5.10.0 on linux: $ARGV = { 'WHEN' => '12:34' }; $string = 'It is thyme';
    Considering that the MSWin32 Perl also gives you what you want, I'd say you encountered a bug in the cygwin port. It may be worth sending in a bug report; preferably with the code fragment turned into a test.

    As for a fix, I've one suggestion. Try turning the my $ARGV into local %ARGV, and use a fully qualified path when refering to $ARGV{WHEN}. But that's just a guess, and since my perl doesn't have the bug in the first place, I cannot test whether it makes a difference.

      I don't think that it's Cygwin - or at least not only Cygwin. Here are my results on OS X. The 5.8 is the system Perl, and the 5.10 is my own:

      hektor ~ $ ./retest Perl 5.10.0 on darwin: $ARGV = { 'WHEN' => undef }; $string = 'It is thyme'; hektor ~ $ /usr/bin/perl retest Perl 5.8.8 on darwin: $ARGV = { 'WHEN' => '12:34' }; $string = 'It is thyme';

      A quick second look: I'm now thinking it has something to do with the configuration. These two are from the same machine (Debian Lenny), but they're both 5.10.0. Nevertheless, the results differ. The first 5.10.0 is once again my own build, but the second is Debian's. (I applied a few patches to my own, particularly one to stop a regular expression related memory leak, but I believe that Debian has that patch as well - and more.)

      telemachus ~ $ perl testre Perl 5.10.0 on linux: $ARGV = { 'WHEN' => undef }; $string = 'It is thyme'; telemachus ~ $ /usr/bin/perl testre Perl 5.10.0 on linux: $ARGV = { 'WHEN' => '12:34' }; $string = 'It is thyme';

        Thank you all for your help and insight. Unfortunately, I see little recourse but to abandon my use of Getopt::Euclid. I'm not going to rewrite existing scripts for now, but I won't use Getopt::Euclid for a new research project I am starting.

        The straw that broke the camel's back (so to speak) is the inconsistent failures. It works for me on OS X with perl 5.10 but fails for telemachus. Further, it works with one build of perl 5.10.0 on linux and fails for a different build of 5.10.0 on the same system.

        I lack time and motivation to trace the failure all the way back to perl build configuration settings, especially since it appears that Damian Conway has abandoned Getopt::Euclid.

        Once again, thank you for your help.


        P.S. I rewrote my sample code to be a proper Test::More test script. Furthermore, I also reverted to the original regular expression, in case my simplification introduced its own errors.

        That code, when run twice on my system, reports:

Re: Getopt::Euclid victim of 5.10 upgrade
by parv (Priest) on Feb 13, 2009 at 06:34 UTC
    I am curious how would changing to ActiveState Perl change things. Is it that ActiveState provide fixes (for your situation) for the Getopt::Euclid module?
      It works with cygwin perl 5.8. It works with ActiveState 5.8 and 5.10. It works with all versions I use on OS X. It is only cygwin's default perl, version 5.10, that has revealed the 'experimental' regex sensitivity.
        I don't think this is related to the "experimental" nature of (?{ }) either.
      It's been shown to work in linux and earlier version, so I don't think so. I think something broke in cygwin.

Log In?

What's my password?
Create A New User
Node Status?
node history
Node Type: perlquestion [id://743501]
Approved by GrandFather
Front-paged by Narveson
and all is quiet...

How do I use this? | Other CB clients
Other Users?
Others about the Monastery: (6)
As of 2017-08-20 09:52 GMT
Find Nodes?
    Voting Booth?
    Who is your favorite scientist and why?

    Results (313 votes). Check out past polls.