When rebuilding (because of flac dependency subslot change), =x11-libs/wxGTK-3.0.5.1 fails two tests. All tests passed when wxGTK was last built, back in July. The test suite reported these assertion failures: /var/tmp/portage/x11-libs/wxGTK-3.0.5.1/work/wxWidgets-3.0.5.1/tests/intl/intltest.cpp:206:Assertion Test name: IntlTestCase::DateTimeFmtFrench equality assertion failed - Expected: %a %d %b %Y %H:%M:%S %Z - Actual : %a %d %b %Y %H:%M:%S - French date and time /var/tmp/portage/x11-libs/wxGTK-3.0.5.1/work/wxWidgets-3.0.5.1/tests/misc/dynamiclib.cpp:76:Assertion Test name: DynamicLibraryTestCase::Load assertion failed - Expression: lib.IsLoaded() Failures !!!
Created attachment 823498 [details] x11-libs:wxGTK-3.0.5.1:20221010-165850.log.gz
Created attachment 823501 [details] emerge-info.txt
Can't repro the DateTime one (we probably just need to set locale) but the DynamicLibrary one does repro on amd64 - trying to find how I missed this. I did all my testing on sparc and I'm positive everything passed there.
(In reply to matoro from comment #3) > Can't repro the DateTime one (we probably just need to set locale) but the > DynamicLibrary one does repro on amd64 - trying to find how I missed this. > I did all my testing on sparc and I'm positive everything passed there. *complete* guess would be that it doesn't like sandbox and LD_PRELOAD ordering affects it.
This is looking even uglier than I thought. I tested on ppc64le out of curiosity, and it does NOT fail the DynamicLibrary test but does have a signed/unsigned char failure! /var/tmp/portage/x11-libs/wxGTK-3.0.5.1/work/wxWidgets-3.0.5.1/tests/strings/vararg.cpp:136:Assertion Test name: VarArgTestCase::CharPrintf equality assertion failed - Expected: value is -16 (int) - Actual : value is 240 (int) Failures !!! Run: 1242 Failure total: 1 Failures: 1 Errors: 0 And this isn't even half of the test suite for 3.0, this is just the ones that can run without a GUI. I didn't even bring in the GUI tests because they didn't work. Shouldn't have brought this in without testing across more architectures. Plus we can't report anything to upstream because they've moved on from 3.0 and the test framework in 3.2 is completely different. Maybe we should just give up, revert and say no tests until we're caught up with upstream? It's pretty bad to be this far behind.
