Created attachment 153575 [details]
Debug log for CppUnitTest_sc_dataprovider_test
I'm running the unit tests on a Tumbleweed ppc64le machine and the
CppUnitTest_sc_dataprovider_test test fails:
ScDataProvidersTest::testCSVImport finished in: 1810ms
[New Thread 0x7fffac59f160 (LWP 22385)]
[Thread 0x7fffac59f160 (LWP 22385) exited]
ScDataProvidersTest::testDataLargerThanDB finished in: 4ms
[New Thread 0x7fffac59f160 (LWP 22386)]
[Thread 0x7fffac59f160 (LWP 22386) exited]
ScDataProvidersTest::testHTMLImport finished in: 4ms
[New Thread 0x7fffac59f160 (LWP 22387)]
[Thread 0x7fffac59f160 (LWP 22387) exited]
ScDataProvidersTest::testXMLImport finished in: 139ms
Thread 1 "cppunittester" received signal SIGSEGV, Segmentation fault.
0x00007fffad6d42bc in ?? ()
#0 0x00007fffad6d42bc in ()
#1 0x00007fffb785b114 in __run_exit_handlers () at /lib64/libc.so.6
#2 0x00007fffb785b308 in exit () at /lib64/libc.so.6
#3 0x00007fffb78382a0 in generic_start_main.isra () at /lib64/libc.so.6
#4 0x00007fffb7838484 in __libc_start_main () at /lib64/libc.so.6
#5 0x0000000000000000 in ()
A debugging session is active.
Inferior 1 [process 22381] will be killed.
Quit anyway? (y or n) [answered Y; input not from terminal]
For me it looks like the actual tests passes, but the unittester segfaults.
Created attachment 153576 [details]
Valgrind log for CppUnitTest_sc_dataprovider_test
yeah, crash on cleanup of globals on exit, I doubt its a specific LibreOffice bug personally.
This looks like some kind of shutdown/C++ destructor problem, probably because the order of destruction of C++ global objects is undefined, and the order slightly different on your machine, and some global object is attempting to access another global object during shutdown, but the accessee is already dead.
How could I/we further debug this? So you assume it's a compiler bug (on ppc64le). Did I understand that correctly?
seems to be working ok in our Fedora rawhide/32 build i.e.
[build CUT] sc_dataprovider
(In reply to Caolán McNamara from comment #2)
> yeah, crash on cleanup of globals on exit, I doubt its a specific
> LibreOffice bug personally.
Does it make sense to ask Jonathan Brielmaier to report the problem to OpenSuse instead ?
(In reply to Caolán McNamara from comment #5)
> seems to be working ok in our Fedora rawhide/32 build i.e.
> [build CUT] sc_dataprovider
Closing as RESOLVED NOTOURBUG