On 05/09/2016 05:44 PM, Helen Koike wrote: > In case of failure, print "q.create_bufs(node, 1, &fmt) != EINVAL" instead > of "ret != EINVAL" > > Signed-off-by: Helen Koike <helen.koike@xxxxxxxxxxxxxxx> > --- > > Hello, > > I was wondering, why the q.create_bufs is expected to should return EINVAL in this test? The height and size are set to half of the original values, and the type and memory doesn't seems to change. For all drivers currently in the kernel the buffer size that create_bufs wants should be >= the size of the current format. This checks if the drivers perform that test correctly. In theory it should be possible to allocate smaller buffers as well and drivers that allow on-the-fly format changes, but such drivers do not exist today. If such drivers arrive, then this test should probably be modified. Regards, Hans > > Thank you > > utils/v4l2-compliance/v4l2-test-buffers.cpp | 3 +-- > 1 file changed, 1 insertion(+), 2 deletions(-) > > diff --git a/utils/v4l2-compliance/v4l2-test-buffers.cpp b/utils/v4l2-compliance/v4l2-test-buffers.cpp > index 6c5ed55..fb14170 100644 > --- a/utils/v4l2-compliance/v4l2-test-buffers.cpp > +++ b/utils/v4l2-compliance/v4l2-test-buffers.cpp > @@ -955,8 +955,7 @@ int testMmap(struct node *node, unsigned frame_count) > fmt.s_height(fmt.g_height() / 2); > for (unsigned p = 0; p < fmt.g_num_planes(); p++) > fmt.s_sizeimage(fmt.g_sizeimage(p) / 2, p); > - ret = q.create_bufs(node, 1, &fmt); > - fail_on_test(ret != EINVAL); > + fail_on_test(q.create_bufs(node, 1, &fmt) != EINVAL); > fail_on_test(testQueryBuf(node, cur_fmt.type, q.g_buffers())); > fmt = cur_fmt; > for (unsigned p = 0; p < fmt.g_num_planes(); p++) > -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
