On Fri, Jun 26, 2015 at 08:55:07PM +0100, Jose Fonseca wrote: > No objections from me (I have enough crashes and fails to worry about, so > "warns" are rarely something I have time for.) > > It might make sense to have an option like GCC's "-Werror" that makes > warnings errors. Though I suppose anybody interesting in finding out > warnings can also run the json backend.
Yeah I guess adding an option to decide the JUnit warn mapping would be useful. At least for igt at least everything dmesg-* (which is just the kernel warning) is considered a failure since we can't crash the test process itself easily with asserts in kernel code. Same holds true for running openl* tests with --dmesg imo. But I still maintain that if your testcase is printing unreliable stuff on stderr, then the testwrapper should just filter that out. In igt we take great care that anything printed to stderr really is a genuine problem. Generally it's stuff like "results not quite as accurate as expected, but not bad enough to require exiting the testcase". This gives us more coverage in case of tests for complex features, where unfortunately different regressions can pile up. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ Piglit mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/piglit
