Eric Wilhelm wrote:
Does that mean we'll have to ship another round of alpha?
I don't think so (see below).
John, is there a test to go with this or is the coverage in the version
dist? Any reason that this would break our fully-tested status?
There are tests in the version.pm package on CPAN:
is $badv, '1.002003', "Deal with badly serialized versions from YAML";
my $badv2 = bless { qv => 1, version => [1,2,3] }, "version";
is $badv2, 'v1.2.3', "Deal with badly serialized versions from YAML ";
If a META.yml was generated with an older release of version.pm (pre
5.10.0), the serialized form lacked an 'original' hash entry (which is
what stringify defaults to). However, between Andreas and myself, we
were only able to identify 41 packages on the entirety of CPAN that had
the issue (and some of those were from non-current package releases).
Maybe I didn't say "code freeze" or just don't have the authority to
match the responsibility which I feel I would be assuming -- I don't
want to feel like I have to drop everything and triage the CPAN because
we missed something, so I take the smoke testing very seriously. Or
maybe I'm just burnt out. Anyway, I don't have any more time for
Module::Build at the moment.
I'm really sorry that I missed sending this; I had the patch sitting in
my checkout and when I went to update my mirror, I discovered that I had
never sent it upstream. If it doesn't get added now, it can added for
the next release; it is really a belt and braces issue more than
anything else. If anyone gets bitten by this, they can install the CPAN
release of version.pm and it will get corrected.
John
