kalarm fails in some testcases on 32-bit platforms like this:
FAIL! : KADateTimeTest::constructors() Compared values are not the same
Actual (datetimeTz.qDateTime()) : 2001/02/13 11:45:14.000[GMT]
Expected (dtLocal.toTimeZone(london)): 2001/02/13 03:45:14.000[GMT]
Loc: [./src/kalarmcalendar/autotests/kadatetimetest.cpp(457)]
In debian/rules this line:
testsuite_failing_archs := armel armhf i386
disables the checks on the mentioned platforms (why were the others like m68k, powerpc and such left out?).
But this just hides a problem I assume.
Does anyone has an idea why those tests fail?
Has someone already analyzed?
Is it related to the time64 transition?
Helge
Hej Helge, Am Samstag, 4. Januar 2025, 16:08:05 MEZ schrieb Helge Deller: I doubt anyone has seriously looked at the test failures. We just don't have the resources to investigate that. The reason why only some archs were explicitly mentioned is because they are release architectures and the package otherwise wouldn't migrate to testing. I guess no one really cared about kalarm on m68k and the likes.
Hello Patrick, Sure, I understand that. I was just interested if somebody already had a generic idea why they fail (and thus for now disabled the testsuites, e.g. because the failure wasn't criticial anyway). My question can then maybe simply be reduced to: If someone disables the testsuite for release archs, then please also disable those for non-release architectures which fail the same way. This would help people taking care of non-release architectures a lot. Maybe I'll do when I find time. It's low priority for me too atm. Thanks! Helge
Back to trying to build kalarm on hppa and I see hppa still hasn't been added to testsuite_failing_archs. I suspect this fails on on 32-bit architectures because the t64 patch for libstdc++-v3 is broken. See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99832. I have pushed Matthias on this issue but haven't had a response. Dave
Can you confirm whether the tests pass on HPPA or some other 32-bit arch with the patch mentioned in gcc bug #99832 ? If so that should definitely be a bug against the gcc package, not kalarm. Ignoring the test failure for a real issue doesn’t look like a good idea… Thanks, -- Aurélien