Open wchristian opened 8 years ago
This appears to be a CPAN Testers issue. The database we're pulling from doesn't seem to be up to date. I've alerted preaction about the issue.
@haarg any follow up on this?
I've poked @preaction about this again.
From @andk on IRC:
the release.db stuff was commented out because it never worked well enough. it again and again produced invalid sqlite dumps. It was extremely frustrating to download 8 GB to get data that blew up after several hours of processing the outcome was that barbie offered CPAN::Testers::WWW::Reports::Query::Reports which works well for David Cantrell and me, we are using two similar scripts, mine is refill-cpanstatsdb.pl in http://repo.or.cz/r/andk-cpan-tools.git
In the meantime, we might have the release.db fixed here at the QAH. Will keep y'all up to date here.
We've re-run the summary, and the release database should be available. Could you run your import and let me know?
I'm seeing results for some reasonably recent releases, so I believe this is working now.
This is broken again.
Alright. I'll take a look hopefully today if not tomorrow. Tracking at https://github.com/cpan-testers/cpantesters-project/issues/9
I've found the appropriate commands to run, and the new release is being created. It's got some catching up to do, so it might be a day or two (or more, but probably less than it was during the QAH). I've added a cron job to run the missing command to copy the right data a couple times a day. In a week or so, I'll check to make sure everything is still working correctly.
So, I guess we can come back to this ticket mid-week and see if things are back to normal.
It's only gotten through 1/5 of the records (1,000,000 in about a day). The cronjob will protect us from this in the future, but it looks like it'll be another 4 days before the release is fully in-sync. I've been stopping the update process nightly to rebuild the sqlite database with the records that have been gotten though, so new(-ish) data should at least be flowing to metacpan.
CPAN Testers has processed up to 72m out of 73.3m, and the release SQLite database has been updated. A test check of metacpan doesn't seem to show the cpantesters data being updated: https://metacpan.org/release/DROLSKY/Params-CheckCompiler-0.07 (471 pass, 0 fail/na/unknown) and https://metacpan.org/release/DROLSKY/Params-CheckCompiler-0.06 (353 pass, 0 fail/na/unknown) are both in the release database, but do not seem to have been updated on the site.
Is Text::CSV_XS still in those 1.3m ?
$ perl -MMetaCPAN::Client -MDP -wE'my$m=MetaCPAN::Client->new (version => "v1")->release({distribution=>"Text-CSV_XS"});while (my $r = $m->next){ DDumper [ $r->name, $r->tests ] if $r->name =~ /1\.2[345]/ }'
[ 'Text-CSV_XS-1.24',
undef
]
[ 'Text-CSV_XS-1.23',
{ fail => 0,
na => 0,
pass => 985,
unknown => 0
}
]
[ 'Text-CSV_XS-1.25',
undef
]
Yes.
sqlite> select * from release where dist="Text-CSV_XS" and version > 1.2;
Text-CSV_XS|1.20|1053|2|0|1
Text-CSV_XS|1.21|685|0|0|0
Text-CSV_XS|1.22|602|15|0|0
Text-CSV_XS|1.23|1011|0|0|0
But 1.23 has a few more PASSes
The Text::CSV_XS appears to be in the release.db now:
sqlite> select * from release where dist="Text-CSV_XS" and version > 1.2;
Text-CSV_XS|1.20|1053|2|0|1
Text-CSV_XS|1.21|685|0|0|0
Text-CSV_XS|1.22|602|15|0|0
Text-CSV_XS|1.23|1011|0|0|0
Text-CSV_XS|1.24|699|0|0|0
Text-CSV_XS|1.25|725|0|0|0
Which matches with what Metacpan has for Text::CSV_XS now. So this appears to be resolved. Until I get a monitor for this process, I will keep checking on it for the next week or so to ensure it stays stable.
Numbers don't seem to be matching again:
https://metacpan.org/release/Throwable-SugarFactory
http://matrix.cpantesters.org/?dist=Throwable-SugarFactory+0.152700
I mentioned in #1870 that the release data doesn't have all the data that the matrix might have: Development Perl versions are not included in the release data, but the matrix probably includes them.
The numbers in the SQLite database Metacpan receives match up with the numbers in the database via the API: http://api.cpantesters.org/v1/release/dist/Throwable-SugarFactory
We could change that, but I'm not sure I like the idea of a development Perl showing up in the test results, since it could be not the author's fault.
The stats for Path::Tiny 0.114 today aren't even close.
CPAN Testers processing is a bit slow today, and likely missed the sync window for MetaCPAN.
It was certainly transient as the figures on metacpan now look much closer to the cpantesters numbers again.
Missing:
https://metacpan.org/pod/Moose
https://metacpan.org/release/Throwable-SugarFactory
Existing:
https://metacpan.org/pod/distribution/DBIx-Class/lib/DBIx/Class.pod