Hi Wookey, and sorry for not noticing this bug earlier (not the failure itself though). On Wed, 24 Dec 2014 19:05:54 +0000, Wookey wrote: > It has failed twice on the autobuilders, although not with exactly the same issue: > > It built OK on a porterbox, according to https://lists.debian.org/debian-wb-team/2014/12/msg00010.html Yes, I was able to build it successfully on a porterbox again, and also the same version was built successfully on Canonical's hardware. Thus I believe that it can be an issue with buildds. I have also raised this on debian-arm@ (see thread starting at [1]), but was not able to get any help from there. > Do the failures below give any clues as to might be wrong? They seem > similar but not identical. Is there parallelism in this build making > it less obvious which test is actually at fault? In all logs, the crashing test is tst_QXmlSchema. The last successful test in tst_QXmlSchema is always tst_QXmlSchema::loadSchemaDataSuccess(), so the failing test is one that comes after it — that is tst_QXmlSchema::loadSchemaDataFail() [2]. Of course I can just disable that particular test on arm64, but that will be hiding an issue, which can actually affect not only this package. If you have access to arm64 hardware and can reproduce this build failure, having a stacktrace will be very useful for me. To run the test manually, cd to tests/auto/xmlschema and run the command that appears in the build logs (one ending with ./tst_qxmlschema). Prepending gdb before ./tst_qxmlschema should work as well. Also, why is this tagged patch? I don't see any patch here :) [1] https://lists.debian.org/20150118122343.9789.68359@mitya57.me [2] http://sources.debian.net/src/qtxmlpatterns-opensource-src/5.4.0-2/tests/auto/qxmlschema/tst_qxmlschema.cpp/#L285 -- Dmitry Shachnev
Attachment:
signature.asc
Description: OpenPGP digital signature