Discussion:
0.54_01 - Missing prerequisites causing failures from CPAN-testers.
s***@public.gmane.org
2014-04-07 04:17:50 UTC
Permalink
Hi,

This is a bit strange.
In 0.54_01 top level Makefile.PL, I've replaced the old code that handled
the conditional loading of {Test::Warn=>$twv} into PREREQ_PM, with a cleaner
rendition. (See attached patch.)

In a nutshell, this new rendition is:

###########################
my $twv = $] > 5.013 ? 0.23 : 0.21;

my $prereq_pm = {
Data::Dumper => 2.09,
Digest::MD5 => 2.09,
Parse::RecDescent => 1.80,
File::Spec => 0.8,
Test::Warn => $twv,
};
delete $prereq_pm->{'Test::Warn'} if $] <= 5.007;

###########################

But now, for the first time, I'm seeing test failures on 5.16.3 because
Test::Warn is not getting loaded. Furthermore, these failing reports specify
*no* prerequisites:

[quote]
------------------------------
PREREQUISITES
------------------------------

Prerequisite modules loaded:

No requirements found

------------------------------
[/quote]

That's rubbish - Parse::RecDescent is *always* a pre-requisite.
I surmise that it must be this revised code that is causing this error to
arise .... though I'm buggered if I can see how that could be.

To date this has happened in only 2 cases - both 5.16.3, from different
testers, different systems:
http://www.cpantesters.org/cpan/report/9f275d62-7864-1014-ad6f-831c93e8ee67
http://www.cpantesters.org/cpan/report/8bdf6702-bdd8-11e3-8a89-eea0f1eb6092

I don't have 5.16.3 installed. Any insights into what's going awry ?
Otherwise I'll just revert it to the old code (whose only problems were its
long-windedness and lack of sophistication).

Of those 2 failing reports, the first comes from Christian Walde.
He also tested 0.54 a few days ago with what looks to me to be the very same
smoker. There were no problems then as regards getting the prereqs set
correctly:

[quote]
------------------------------
PREREQUISITES
------------------------------

Prerequisite modules loaded:

requires:

Module Need Have
------------------- ---- --------
Data::Dumper 2.09 2.139
Digest::MD5 2.09 2.52
File::Spec 0.8 3.40
Parse::RecDescent 1.8 1.967009
Test::Warn 0.23 0.30

build_requires:

Module Need Have
------------------- ---- --------
ExtUtils::MakeMaker 0 6.63_02


------------------------------
[/quote]

That report (which contains a different failure) can be found in full at:
http://www.cpantesters.org/cpan/report/63a42468-770a-1014-9b4d-7b2a93e8ee67

Cheers,
Rob
s***@public.gmane.org
2014-04-09 03:30:35 UTC
Permalink
-----Original Message-----
Post by s***@public.gmane.org
But now, for the first time, I'm seeing test failures on 5.16.3 because
Test::Warn is not getting loaded. Furthermore, these failing reports
specify
Same bug (slightly different error message) now appears in the following
reports from David Cantrell:

Inline-0.54_01:
- x86_64-linux-thread-multi / 5.16.0:
- FAIL
http://www.cpantesters.org/cpan/report/867a7f06-bf39-11e3-8187-2cd48706f0e4

- x86_64-linux-thread-multi / 5.16.1:
- FAIL
http://www.cpantesters.org/cpan/report/844f1a84-bf39-11e3-8d6b-d1d28706f0e4

- x86_64-linux-thread-multi / 5.16.2:
- FAIL
http://www.cpantesters.org/cpan/report/e9e8be7a-bf40-11e3-b294-b5278806f0e4

- x86_64-linux-thread-multi / 5.18.0:
- FAIL
http://www.cpantesters.org/cpan/report/2693b658-bf47-11e3-8a6c-1a168806f0e4

- x86_64-linux-thread-multi / 5.18.1:
- FAIL
http://www.cpantesters.org/cpan/report/fda15166-bf40-11e3-8787-e7c08706f0e4

- x86_64-linux-thread-multi / 5.18.2:
- FAIL
http://www.cpantesters.org/cpan/report/80a93330-bf41-11e3-8a9d-cb0f8806f0e4

Cheers,
Rob
s***@public.gmane.org
2014-04-09 09:15:23 UTC
Permalink
-----Original Message-----
Post by s***@public.gmane.org
But now, for the first time, I'm seeing test failures on 5.16.3 because
Test::Warn is not getting loaded. Furthermore, these failing reports
Turns out there's another 0.54_01 at
http://search.cpan.org/CPAN/authors/id/R/RU/RURBAN/Inline-0.54_01.tar.gz

I've just re-released 0.54_01 as 0.54_03 (no changes) in order to dissociate
it from the other 0.54_01.

(Bumping the version number to 0.54_02 would have achieved nothing as
there's also
http://search.cpan.org/CPAN/authors/id/R/RU/RURBAN/Inline-0.54_01.tar.gz)

Cheers,
Rob
Reini Urban
2014-04-09 14:41:23 UTC
Permalink
Post by s***@public.gmane.org
This is a bit strange.
In 0.54_01 top level Makefile.PL, I've replaced the old code that
handled the conditional loading of {Test::Warn=>$twv} into PREREQ_PM,
with a cleaner rendition. (See attached patch.)
###########################
my $twv = $] > 5.013 ? 0.23 : 0.21;
my $prereq_pm = {
Data::Dumper => 2.09,
Digest::MD5 => 2.09,
Parse::RecDescent => 1.80,
File::Spec => 0.8,
Test::Warn => $twv,
};
delete $prereq_pm->{'Test::Warn'} if $] <= 5.007;
I'm also baffled why this fails now. Something like works everywhere
else for me. And it tested ok on all my perls.

But since it's just an aesthetical change to reduce code
feel free to revert it. Sorry.
--
Reini

Working towards a true Modern Perl.
Slim, functional, unbloated, compile-time optimizable
s***@public.gmane.org
2014-04-10 23:34:57 UTC
Permalink
-----Original Message-----
From: Reini Urban
Sent: Thursday, April 10, 2014 12:41 AM
To: inline-***@public.gmane.org
Subject: Re: 0.54_01 - Missing prerequisites causing failures from
CPAN-testers.
Post by s***@public.gmane.org
###########################
my $twv = $] > 5.013 ? 0.23 : 0.21;
my $prereq_pm = {
Data::Dumper => 2.09,
Digest::MD5 => 2.09,
Parse::RecDescent => 1.80,
File::Spec => 0.8,
Test::Warn => $twv,
};
delete $prereq_pm->{'Test::Warn'} if $] <= 5.007;
I'm also baffled why this fails now. Something like works everywhere else
for me. And it tested ok on all my perls.
There was some META_MERGE data also added to the Makefile.PL, and no
META.yml or META.json included in the distro.
I've just uploaded 0.54_01 yesterday, which contains MakeMaker-generated
META.yml and META.json.

0,54_01 is looking good - Christian Walde's 5.16.3 Windows smoker (which was
one of the "offending" machines) is back to normal.
But since it's just an aesthetical change to reduce code
feel free to revert it. Sorry.
That's ok. It's a worthwhile change ... and I'm now thinking it's probably
not even that change that triggered the problem. (More likely the META_MERGE
stuff in conjunction with the absence of META.json and META.yml.)

Cheers,
Rob
s***@public.gmane.org
2014-04-11 00:57:57 UTC
Permalink
-----Original Message-----
Post by s***@public.gmane.org
There was some META_MERGE data also added to the Makefile.PL, and no
META.yml or META.json included in the distro.
I've just uploaded 0.54_01 yesterday, which contains MakeMaker-generated
META.yml and META.json.
0,54_01 is looking good - Christian Walde's 5.16.3 Windows smoker (which
was one of the "offending" machines) is back to normal.
I meant 0.54_04 ... not 0.54_01.

Groans,
Rob

Loading...