So probably this will happen after every new release --- only my smokers (15 VMs with up to ten perl versions each) will run the test suite more than hundred times shortly after the release, and there are other smoker systems, too.
At least I think the diagnostics could be improved. Currently looking at the test log a user cannot guess that the failures are actually harmless. And I think the test scripts could act better if this situation is encountered, e.g. by skipping the rest of the tests (I did something similar in the test suite of Geo-Coder-Googlev3).
After the new release the first reports from CPAN Testers looked good, but then it turned red again: http://fast-matrix.cpantesters.org/?dist=WWW-Tumblr%205.3;reports=1#sl=7,1
Looking with a HTTP logger the reason seems to be:
So probably this will happen after every new release --- only my smokers (15 VMs with up to ten perl versions each) will run the test suite more than hundred times shortly after the release, and there are other smoker systems, too.
At least I think the diagnostics could be improved. Currently looking at the test log a user cannot guess that the failures are actually harmless. And I think the test scripts could act better if this situation is encountered, e.g. by skipping the rest of the tests (I did something similar in the test suite of Geo-Coder-Googlev3).