If you specify to generate xml outputs and use the generated files in a build
server like hudson, problems occur if a test binary segfaults. Currently no
output is written then and hudson has no chance to see if the test was executed
but failed or not, because there simply is no xml file generated. The absence
of a xml file should not be the indication for a crashed test.
A simple solution for this problem would be to first write an xml file
indicating a failure of the test program, then execute all unit tests, and in
the end replace the contents of the xml file with the real results. If the unit
tests segfault in between, an error file is still left over.
Original issue reported on code.google.com by johannes...@googlemail.com on 6 Dec 2010 at 5:02
Original issue reported on code.google.com by
johannes...@googlemail.com
on 6 Dec 2010 at 5:02