IMHO either tests should fail on older patch release or an appropriate skip of the test should be done if the version is too old. I am not in favor of the approach in the patch which actually "hides" the issue as Christophe pointed out. I am a little bit in favor of having it fail as it makes clear what stuff isn't fixed in the old release.
Regards Rüdiger C2 General Von: Eric Covener <[email protected]> Gesendet: Donnerstag, 19. Dezember 2019 23:13 An: Apache HTTP Server Development List <[email protected]> Cc: pgajdos <[email protected]> Betreff: Re: t/modules/{brotli,deflate}.t: handle the case < 2.4.42 On Thu, Dec 19, 2019 at 4:32 PM Marion & Christophe JAILLET <[email protected]<mailto:[email protected]>> wrote: Hi, I've received the attached patch against the test framework. This is not the first time Petr propose such patches and most (or maybe even all) have been merged. The goal is to avoid new warnings on suse tests when a recent test framework is used on an older version of httpd. However, this particular case puzzles me. I see r1868313 as a fix on a non-RFC compliant behavior. So I'm not sure this should be "hidden" only because older versions are known to fail. In other word, what is the "semantic" behind the test framework? - Testing against regression only? (in this case including Petr's patch makes sense to me) - or - - Checking that the behavior is correct (in this case, I'm not fan of "hiding" things that are known to fail. A failure would mean: "you should upgrade to have this fixed") I think the tests should either skip or have changed expectations when run against an old build, not fail. In other words the tests are for us, not end users curious how correct what they have is.
