Brendan McGrath (@redmcg) commented about dlls/mf/tests/transform.c:
+ { + .length = extra_width * actual_height * 4, + .compare = compare_rgb32, .compare_rect = {.top = 12, .right = 82, .bottom = 96}, + .dump = dump_rgb32, .size = {.cx = extra_width, .cy = actual_height}, + }; + const struct sample_desc rgb32_extra_width_sample_desc = + { + .attributes = output_sample_attributes, + .sample_time = 0, .sample_duration = 10000000, + .total_length = actual_width * actual_height * 4, + .buffer_count = 1, .buffers = &rgb32_extra_width_buffer_desc, + }; + const struct buffer_desc rgb32_extra_height_buffer_desc = + { + .length = actual_width * extra_height * 4, + .compare = compare_rgb32, .compare_rect = {.top = 12, .right = 82, .bottom = 96},
.compare = compare_rgb32, .compare_rect = {.top = 20, .right = 82, .bottom = 104},
This is my bad, but I was surprised to find that Windows letter boxes the image when the output is taller, and I realise now that I never updated the `compare_rect` to reflect that. I think top should now be 20 and bottom 104 (but I haven't had a chance to test that). To be completely honest, I'm not against removing the vertical tests for now, as I never included tests for going to a shorter buffer either. That would leave this MR to focus on stride (and a future MR for vertical tests). But I'm not strongly opinionated either way. And to be a complete PITA, I'll be mostly unavailable the next few days. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/6220#note_78076