On 12/10/17 15:39, Chris Wilson wrote:
Quoting Daniele Ceraolo Spurio (2017-10-12 23:30:40)
We already had coverage in prime_vgem.c, but since we're testing
pread/pwrite ioctl it makes more sense to have it in the corresponding
test binaries
Redundancy is not an issue (sometimes it is a virtue, the slight
variations in test setup lead to wider coverage and serendipity - when we
don't have a fuzzer, we have to rely on random variations between
authors). The purpose of each test is different, one is testing prime and
interactions with i915 on/from a foreign bo; the other looking at
pwrite/pread in depth. Each subtest is meaningful as part of the
encompassing test.
The counter proposal is to be able to select a cross-selection of
subtests for a given set of tags/functions/code. And we still need a
fuzzer, model based or not.
-Chris
Would it be ok to just drop patch 4/4 and rephrase the commit messages?
There is some extra redundancy compared to before as I've added
read/write checks to pread/pwrite but that shouldn't hurt IMHO and I'd
prefer to keep it.
Thanks,
Daniele
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
