On Fri, Nov 15, 2013 at 06:08:27PM +0100, Daniel Vetter wrote: > On Fri, Nov 15, 2013 at 04:33:27PM +0000, Damien Lespiau wrote: > > Tests tend to assume they are the only thing running, some need DRM > > master, some fill the memory. All in all, they are some correctness > > issues if we run tests concurrently. > > > > Acked-by: Ben Widawsky <ben@xxxxxxxxxxxx> > > Signed-off-by: Damien Lespiau <damien.lespiau@xxxxxxxxx> > > Concurrency is a per-test opt-in thing in piglit. So this is cargo-cult > and worse will prevent us to run a subset of tests to run in parallel > (e.g. once we have a naming scheme for abi stuff). Right now, the concurrency runs have big problems. Running anything along drv_module_reload will fail to open the device (and as it fails quickly, a lot of tests will end up being "skipped" (instead of failing, which is another bug)). Hence using --no-concurrency for now. It's something we can maintain in tree though, and remove once we are more confident it has a reasonable chance to work. -- Damien _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx
