Bug #56796 for version: version.pm error detection fail

Wed Apr 21 16:09:19 2010 zefram [...] fysh.org - Ticket created

Subject:	version.pm error detection fail
Date:	Wed, 21 Apr 2010 21:09:01 +0100
To:	bug-version [...] rt.cpan.org
From:	Zefram <zefram [...] fysh.org>

On Perl 5.11.0 to 5.11.3 inclusive, version.pm 0.82 fails its test suite: $ ./Build test t/01base.t ..... 1/? # Failed test 'Invalid version format (non-numeric data)' # at t/coretests.pm line 68. # '' # doesn't match '(?-xism:non-numeric data)' [many similar test failures omitted] Test Summary Report ------------------- t/01base.t (Wstat: 3072 Tests: 527 Failed: 12) Failed tests: 13-15, 144-146, 275-277, 406-408 Non-zero exit status: 12 t/02derived.t (Wstat: 3072 Tests: 539 Failed: 12) Failed tests: 19-21, 151-153, 283-285, 415-417 Non-zero exit status: 12 t/03require.t (Wstat: 1536 Tests: 266 Failed: 6) Failed tests: 16-18, 147-149 Non-zero exit status: 6 Files=3, Tests=1332, 2 wallclock secs ( 0.32 usr 0.03 sys + 1.07 cusr 0.10 csys = 1.52 CPU) Result: FAIL Failed 3/3 test programs. 30/1332 subtests failed. -zefram

Wed Apr 21 17:15:49 2010 jpeacock [...] cpan.org - Correspondence added

Not a bug. I don't have the time/energy to tweak the #defines for every unreleased Perl snapshot. I can only keep up with the main releases... Sorry John

Wed Apr 21 17:15:56 2010 The RT System itself - Status changed from 'new' to 'open'

Wed Apr 21 17:16:01 2010 jpeacock [...] cpan.org - Status changed from 'open' to 'rejected'

Thu Apr 22 03:03:45 2010 zefram [...] fysh.org - Correspondence added

Subject:	Re: [rt.cpan.org #56796] version.pm error detection fail
Date:	Thu, 22 Apr 2010 08:03:28 +0100
To:	John Peacock via RT <bug-version [...] rt.cpan.org>
From:	Zefram <zefram [...] fysh.org>

John Peacock via RT wrote: Show quoted text

>Not a bug. I don't have the time/energy to tweak the #defines for every >unreleased Perl snapshot.

In order to debug other modules, it's convenient to keep at least the toolchain modules working for dev releases. If I provide a patch, would you accept it? -zefram

Thu Apr 22 03:03:47 2010 The RT System itself - Status changed from 'rejected' to 'open'

Thu Apr 22 08:29:02 2010 john.peacock [...] havurah-software.org - Correspondence added

Subject:	Re: [rt.cpan.org #56796] version.pm error detection fail
Date:	Thu, 22 Apr 2010 08:28:45 -0400
To:	bug-version [...] rt.cpan.org
From:	John Peacock <john.peacock [...] havurah-software.org>

Zefram via RT wrote: Show quoted text

> In order to debug other modules, it's convenient to keep at least the > toolchain modules working for dev releases.

Forward compatible for 5.13.0 maybe, but I don't understand why backporting to 5.11.x is valuable now that 5.12.0 is out. Since version.pm was significantly changed during the 5.11.x timeframe, figuring out which things changed when is not going to be easy. The maze of #define's is there specifically to upgrade certain functions/methods for specific releases (by actually #define'ing the function call to a new name). Figuring out which one is needed where is a pain, which is why I only do it to match public/non-dev releases. Show quoted text

> If I provide a patch, would you accept it?

You can give it a shot if you like, but I'll be upfront that I'm not going to push a new release to CPAN just to maintain compatibility with stale dev releases. I do have some bugfixes that we didn't put into 5.12.0 code (mostly minor leaks), so there will be a catch-up CPAN release soonish (though I'm not making any promises)... John

Thu Apr 22 08:49:09 2010 zefram [...] fysh.org - Correspondence added

Subject:	Re: [rt.cpan.org #56796] version.pm error detection fail
Date:	Thu, 22 Apr 2010 13:48:48 +0100
To:	John Peacock via RT <bug-version [...] rt.cpan.org>
From:	Zefram <zefram [...] fysh.org>

John Peacock via RT wrote: Show quoted text

>Forward compatible for 5.13.0 maybe, but I don't understand why >backporting to 5.11.x is valuable now that 5.12.0 is out.

Right now there's quite a bit of activity around TryCatch, Devel::Declare, and related modules. TryCatch doesn't work on 5.12, and it's turned out to be tough to figure out why. Quite a lot of the clues available come from people trying intermediate versions in the 5.11 series. It definitely doesn't work on 5.11.2 and later. There's a different failure mode on 5.11.0 and 5.11.1, possibly due to prerequisites. Devel::Declare necessarily uses a different strategy on 5.11.2 and later than on earlier versions, due to having been broken by the sub lookup change in 5.11.2, so there's a suspicion that that's related. If we didn't have a working toolchain and supporting modules on 5.11.*, we'd lack these clues and have a much worse time of debugging TryCatch. The 5.11.* releases are of immense value here in debugging a problem relating to 5.12. (Before you ask: this problem was known about before 5.12 was released; there's been a lengthy process of debugging Devel::Declare and *its* supporting modules before we could get this far. The present activity is essentially a continuation of work from the 5.11 cycle.) The TryCatch stuff doesn't (currently) depend on having a recent version.pm working on 5.11. (Later toolchain updates conceivably might make TryCatch indirectly depend on recent version.pm, of course.) It's an illustration of the kind of situation where dev releases have been useful. Looking further back, I've also found the 5.9 releases helpful when porting new modules to older Perl versions: the gradual transition makes it easier to tackle the differences between 5.10 and 5.8. (I don't routinely test modules against 5.9, though, whereas I do routinely test against 5.11. The release quality is much higher with 5.11.) I'll see about working up a patch. No worries about turnaround time. -zefram

Thu Apr 22 13:27:24 2010 john.peacock [...] havurah-software.org - Correspondence added

Subject:	Re: [rt.cpan.org #56796] version.pm error detection fail
Date:	Thu, 22 Apr 2010 13:27:06 -0400
To:	bug-version [...] rt.cpan.org
From:	John Peacock <john.peacock [...] havurah-software.org>

Zefram via RT wrote: Show quoted text

> If we didn't have a working toolchain and supporting modules on 5.11.*, > we'd lack these clues and have a much worse time of debugging TryCatch. > The 5.11.* releases are of immense value here in debugging a problem > relating to 5.12. (Before you ask: this problem was known about > before 5.12 was released; there's been a lengthy process of debugging > Devel::Declare and *its* supporting modules before we could get this far. > The present activity is essentially a continuation of work from the > 5.11 cycle.)

OK, that makes more sense. I think if you treat 5.11.x as if it were 5.10.1, that will DTRT (in essence because it will use the alternate function names to prevent the core ones from being called). NOTE you will have to update the #defines in both vutil.h and vutil.c. In the highly unlikely chance I have some spare time this weekend, I'll take a look at it myself... John

Thu Apr 22 20:29:59 2010 john.peacock [...] havurah-software.org - Correspondence added

Subject:	Re: [rt.cpan.org #56796] version.pm error detection fail
Date:	Thu, 22 Apr 2010 20:30:42 -0400
To:	bug-version [...] rt.cpan.org
From:	John Peacock <john.peacock [...] havurah-software.org>

On 04/22/2010 08:49 AM, Zefram via RT wrote: Show quoted text

> Queue: version > Ticket<URL: https://rt.cpan.org/Ticket/Display.html?id=56796> > > John Peacock via RT wrote:

>> Forward compatible for 5.13.0 maybe, but I don't understand why >> backporting to 5.11.x is valuable now that 5.12.0 is out.

Attached is a patch to make version.pm compile and work under 5.11.0 and 5.11.1 (I had some time waiting for my son's concert to start). I can't say when an official release will be out, but you can point people at this patch in the meantime. John

Message body is not shown because sender requested not to inline it.

Mon Apr 26 15:36:07 2010 zefram [...] fysh.org - Correspondence added

Subject:	Re: [rt.cpan.org #56796] version.pm error detection fail
Date:	Mon, 26 Apr 2010 20:35:48 +0100
To:	John Peacock via RT <bug-version [...] rt.cpan.org>
From:	Zefram <zefram [...] fysh.org>

Attached is a patch, based on version-0.82, to make it compile and pass tests under all 5.11 releases. As a bonus, I tried it out on 5.9 too, which I don't routinely test against. Unmodified version-0.82 turns out to fail on all 5.9 releases, and I made it work on 5.9.{3..5} (also included in this patch). On 5.9.{0..2} there's more that's wrong with it, and I don't think it's worth the effort to debug it there, because those versions were real lemons. -zefram

Message body is not shown because sender requested not to inline it.

Sun May 02 10:18:43 2010 jpeacock [...] cpan.org - Correspondence added

Thanks! Applied.

Sun May 02 10:18:45 2010 jpeacock [...] cpan.org - Status changed from 'open' to 'resolved'