I was looking into why a failed unit test doesn't fail the build and I noticed that we have been failing a unit test on bamboo.openx.org for a while now (circa r2091).
<span class="fail">Fail</span>: ../tests/core/ArchiveProcessing.test.php -> Test_Piwik_ArchiveProcessing -> test_init_today -> Equal expectation fails because 1276803074 differs from 0 by 1276803074 at line 107
interesting, the test is not failing on my box. Definitely something to investigate
It's failing on my ci server too, but not when running the test directly in the browser. Difference is php-cli vs mod_php. I'll take a look at the phpinfo() differences.
(In [2329]) refs #1433 - refactoring
Ok. It was due to these changes.
!isArchivingEnabled() isn't functionally equivalent to isBrowserTriggerArchivingEnabled() because isArchivingEnabled() has a check for isPhpCliMode(). This is what trips up the unit test on the continuous integration server (it uses php-cli to run the unit tests).
(In [2331]) fixes #1433
(In [2340]) Refs #1433