Robert Elz wrote: > I noticed that the b5 mixerctl tests fail - seems to be because the > qemu kernel the tests are run under has no audio (and hence any open > of /dev/mixer* fails). That's fine, and I can (and am in the process > of) cause the tests that will fail to simply be skipped when running > under qemu.
The criterion for skipping the tests should be "no audio", not "running under qemu" - otherwise they will still fail on real hardware that lacks audio devices. > But one of the tests just runs > mixerctl > and expects to see the usage message. In a "normal" system that's > what happens, but it only happens after /dev/mixer (or whatever) has > been opened - which fails under qemu, causing the test to fail. > > Is the general feeling that this is OK, and this test (which doesn't really > need the mixer at all) should also be skipped on qemu, or should mixerctl > be fixed to detect this case and give the usage without attempting to > open the mixer device? I think mixerctl should be fixed, at least for the case where there are no arguments. I don't have an opinion on the "mixerctl -d /dev/mixer2" case. Looks like the problem doesn't occur on amd64 because it has "pseudo-device pad" in GENERIC. I wonder why i386 doesn't. -- Andreas Gustafsson, g...@gson.org
