Bug #50533 for Filter-Crypto: Minimum version of Pod::Usage should be 1.33

Thu Oct 15 15:10:02 2009 ANDK [...] cpan.org - Ticket created

Subject:

Minimum version of Pod::Usage should be 1.33

Thanks to the fact that Pod::Usage is now declared as a prerequisite we get statistics about the versions of Pod::Usage for free. This reveals that all currently 11 FAIL reports in the 5.6.2 area were using Pod::Usage 1.14 which had no working "initialize" method. All PASSes use 1.33 or higher. I'd recommend bumping the version requirement to 1.33. After that there are currently only 2 FAILs not explained, 5551758 and 5550336. At the moment I cannot see a pattern in the two.

Thu Oct 15 19:35:24 2009 shay [...] cpan.org - Correspondence added

On Thu Oct 15 15:10:02 2009, ANDK wrote: Show quoted text

> Thanks to the fact that Pod::Usage is now declared as a prerequisite we > get statistics about the versions of Pod::Usage for free. This reveals > that all currently 11 FAIL reports in the 5.6.2 area were using > Pod::Usage 1.14 which had no working "initialize" method. All PASSes use > 1.33 or higher. > > I'd recommend bumping the version requirement to 1.33. > > After that there are currently only 2 FAILs not explained, 5551758 and > 5550336. At the moment I cannot see a pattern in the two.

I just tried running the failing test script (t/03_script.t) using perl 5.6.2 (which contains Pod::Usage 1.14), and it works fine. I've also gone through all the PodParser / Pod-Parser distributions on CPAN and BackPAN [where have the BackPAN links gone from search.cpan.org, btw?] which contain a Pod::Usage 1.14 (namely distributions 1.18, 1.20 and 1.21) and tried with all of them too, and it still works with all of them. Their own test suites work too, so I can't figure out what's up with the initialize() problem. I'll continue looking into this. I may try blindly adding a dependency on Pod::Usage 1.33 as you suggest, but it would be nice to understand what the problem is.

Thu Oct 15 19:35:25 2009 The RT System itself - Status changed from 'new' to 'open'

Fri Oct 16 03:37:25 2009 ANDK [...] cpan.org - Correspondence added

Thanks for following up on this. Of course, statistics only gives hints and this one was probably not helpful. I did not try old Pod::Usage versions myself because my last remaining 5.6.2 was built with debugging on and this is rejected by Filter::Crypto. Trying to build a new 5.6.2 fails with much bigger problems for me that I haven't tracked down. What's irritating me is the error message I find on the FAILs saying Can't locate object method "initialize" via package "Pod::Usage" (perhaps you forgot to load "Pod::Usage"?) at /home/ct/test/perl-5.6.2/lib/5.6.2/Pod/Usage.pm line 531. Note, that all say line 531. I see in the sources of Pod::Usage 1.14 in line 531: $self->initialize(); And there is no initialize method. But the inheritance tree seems to be Pod::Text => Pod::Select => Pod::Parser => Exporter if I read the code correctly. initialize() in this Pod::Parser is empty. So indeed it should succeed. Getting nowhere:(

Sat Oct 17 21:44:24 2009 shay [...] cpan.org - Correspondence added

On Fri Oct 16 03:37:25 2009, ANDK wrote: Show quoted text

> Thanks for following up on this. > > Of course, statistics only gives hints and this one was probably not > helpful. I did not try old Pod::Usage versions myself because my last > remaining 5.6.2 was built with debugging on and this is rejected by > Filter::Crypto. Trying to build a new 5.6.2 fails with much bigger > problems for me that I haven't tracked down. > > What's irritating me is the error message I find on the FAILs saying > > Can't locate object method "initialize" via package "Pod::Usage" > (perhaps you forgot to load "Pod::Usage"?) at > /home/ct/test/perl-5.6.2/lib/5.6.2/Pod/Usage.pm line 531. >

Yes, it's irritating me too. I've just uploaded a new version (1.26) containing a change to the usage of the blib module in the test scripts which might have some effect on the FAILs, and also specifying a minimum Pod::Usage version of 1.15 because I found that earlier versions didn't always locate perldoc correctly, which caused a test failure when using a Perl with an architecture-dependent lib folder. I'll leave this ticket open for now because I'm not convinced that these changes will necessarily resolve the problem. Regarding your 5.6.2 built with debugging, were you aware that you can run Filter-Crypto's Makefile.PL with a --debug-mode option? That disables the checks for a debugging perl. It also causes a lot of debug output to be emitted by Filter-Crypto, which makes the tests very noisy, but they do still pass.

Mon Mar 26 17:43:23 2012 shay [...] cpan.org - Correspondence added

The last problem with perldoc that I'm aware of (namely, it being called something like perldoc5.15.7 rather than perldoc in configurations with $Config{versiononly} set) was fixed in Filter-Crypto-1.36, so I'm closing this ticket now. It would still be better if Pod::Usage handled things like perldoc5.15.7 better than it currenly does, but I have a workaround in place in Filter-Crypto-1.36 and can live with that for now. If and when Pod::Usage is enhanced as I have requested in CPAN RT#75598 then I will update the minimum version accordingly.

Mon Mar 26 17:43:24 2012 shay [...] cpan.org - Status changed from 'open' to 'resolved'