22. 07. 2025. u 14:13, James E Keenan piše:
[...]
Seven days later, http://fast-matrix.cpantesters.org/?dist=Socket
remains stuck at 2.039 but also displays the advisory "NOTE: no report
for latest version 2.040". I don't know whether this applies to
reports for other CPAN distributions.
So my impression is that our recovery from the problems of July 15 is
incomplete. Could these be investigated? Thank you very much.
Hello James,
short answer: of the three matrix services just use
*https://fast2-matrix.cpantesters.org* . It is the most reliable, it is
"fast" in the sense that uploaded reports are typically listed within
5-10 minutes and it is currently also the matrix variant which is linked
from MetaCPAN. It is not perfect: in rare cases it can respond with a
50x error, but typically the next reload should be successful.
Why fast-matrix (without the "2") still exists? There are very rare
situations (really rare, I think it happened only once since
fast2-matrix was setup) when it can happen that neither fast2-matrix nor
matrix get fresh data, but the metabase log.txt still works, and thus
fast-matrix.
And why fast-matrix was broken? Well, this seems always to be the case
if there's an MetaCPAN outage. All CPAN distributions uploaded during
such a period are /permanently/ missing in some database which log.txt
relies on (Doug, maybe you can tell more about it?). Some minutes ago I
produced a test report for Socket-2.040 and the log.txt line looks like
this:
[2025-07-22T19:26:58Z] [Slaven Rezić] [pass] [/] [x86_64-linux]
[perl-v5.42.0] [d5807482-6731-11f0-a6e5-7e026e8775ea] [2025-07-22T19:26:58Z]
The field [/] should normally show the author/distribution name. Without
it, the log.txt parser for fast-matrix can only drop this report.
Sorry for the confusion,
Slaven
