Uploading 0.54_03 has revealed that the problem in detecting the pre-requisites does, in fact, pertain to *my* release.
Reini’s 0.54_02 doesn’t suffer these failures – so let’s do a diff on his
Makefile.PL files (0.54_02) versus mine (0.54_03).
Diffs are attached.
The C/Makefile.PL difference pertains only to whitespace inside a print()
statement ... surely nothing there that could have an effect.
And the top level Makefile.PL difference is:
##############################################
--- Inline-0.54_02/Makefile.PL 2014-04-05 02:30:50 +1100
+++ Inline-0.54_03/Makefile.PL 2014-04-05 11:54:32 +1100
@@ -54,6 +54,16 @@
VERSION_FROM => 'Inline.pm',
PREREQ_PM => $prereq_pm,
clean => {FILES => '_Inline_test _Inline .Inline'},
+ META_MERGE => {
+ 'meta-spec' => { version => 2 },
+ resources => {
+ repository => {
+ type => 'git',
+ url => 'https://github.com/ingydotnet/inline-pm.git',
+ web => 'https://github.com/ingydotnet/inline-pm',
+ },
+ },
+ },
);
if ($^O =~ /Win32/i) {
##############################################
I don’t know the purpose of that META_MERGE stuff.
It was a patch sent to me by someone – looked benign enough and didn’t pose any
problems on any of my systems, so I applied it.
Is that the source of these FAIL reports ?
If there’s something wrong with that META_MERGE stuff, I’ll fix it.
Otherwise I think I’ll just take no action and ignore these FAIL reports ....
unless someone has a better idea.
Cheers,
Rob
From: [email protected]
Sent: Thursday, April 10, 2014 12:02 PM
To: Barbie ; David Golden
Cc: CPAN Testers Discuss
Subject: Re: Baffling Inline-0.54_01 test failures.
Thanks for the reply, Barbie.
For now, I’ve just uploaded Inline-0.54_03.tar.gz to CPAN, which will ensure
that the reports I receive relate solely to the distro that I’ve uploaded ...
for as long as no-one else uploads an Inline-0.54_03.tar.gz to CPAN, anyway :-)
Cheers,
Rob
From: Barbie
Sent: Thursday, April 10, 2014 9:13 AM
To: David Golden
Cc: Sisyphus ; CPAN Testers Discuss
Subject: Re: Baffling Inline-0.54_01 test failures.
This is one area I would like to improve with the CPAN Testers Reports site. I
often get questions about distributions with the same name, but due the legacy
way this was handled, it's never been adequately resolved, such that older
distributions couldn't be distinguished from each other (at least not
reliably). Now that we have the metabase facts, we can.
Once a few other projects out the way, I will take another look at this, and
see what changes are necessary to use the AUTHOR as well as the version to
differentiate between distributions and associated reports.
Thanks,
Barbie.
--
Birmingham.pm - http://birmingham.pm.org
CPAN Testers - http://cpantesters.org
YAPC Surveys - http://yapc-surveys.org
Perl Jam - http://perljam.info
On Wed, Apr 9, 2014 at 2:50 PM, David Golden <[email protected]> wrote:
On Wed, Apr 9, 2014 at 2:49 AM, <[email protected]> wrote:
> Can I be sure that the tester reports I quoted were in response to the
> latter ? .... or could they have been in response to the former ?
Look at the "Dear XXXX" part of the email, as that is generated from
the author's PAUSE ID.
This is one of the big, historical problems in the CPAN ecosystem.
Many, many services assume that a distribution name is unique and
there was never any restriction on that.
Recently, at the QA Hackathon, PAUSE now will not index a distribution
tarball unless the uploader has permissions on a package corresponding
to the distribution name, but that only keeps the problem out of the
index file. Many smokers test all uploads regardless of indexing
status.
The Metabase that holds the CPAN Testers info actually captures
AUTHOR/TARBALL, but that's not exposed downstream so the "Dear XXXX"
is the best you've got.
David
--
David Golden <[email protected]> Twitter/IRC: @xdg
CMakefile.PL.urban.diff
Description: Binary data
Makefile.PL.urban.diff
Description: Binary data
