Dan Kegel schrieb:
On Fri, Aug 29, 2008 at 1:42 PM, Rico Schüller kgbricola@web.de wrote:
some days ago I've send a patch which broke some other tests, see http://www.kegel.com/wine/patchwatcher/results/274.txt . But this test got a "conformance test ok". I think patchwatcher should also mark patches as bad when they let crash other tests.
The full run log from that patch is at http://kegel.com/wine/patchwatcher/results/274.log Which of the test failures in there should have triggered a 'bad patch' result?
patchwatcher does a full 'make test' and looks for new test failures, so it's doing what you ask. The only wrinkle is that it has a blacklist of flaky tests, see http://code.google.com/p/winezeug/source/browse/trunk/patchwatcher/patchwatc... Is the test failure you're talking about on that blacklist? If so, patchwatcher will ignore it, as I've found the tests to be unreliable.
It's possible that the low-end nvidia motherboard graphics on the machine running patchwatcher contributed to this flakyness. I'm moving patchwatcher to a faster machine with a better nvidia card at the moment, perhaps that will let me shorten the blacklist.
- Dan
My patch let the ddraw:d3d.c crash on wine (search for "0x60e52ad9 IWineD3DBaseSwapChainImpl_GetBackBuffer" in the log (http://kegel.com/wine/patchwatcher/results/274.log)). So it isn't in the blacklist and shouldn't be there. The problem here is that my patch breaks the test (the test crashes) but the patchwatcher didn't recognize it.
Cheers Rico