On Wed Feb 11 19:28:06 2026 +0000, Elizabeth Figura wrote:
My sincere apologies for dropping this on the ground. I wanted to prioritize regressions for regression season, and then ended up forgetting about this altogether. On retesting I see these failures with unmodified upstream wine: transform.c:5022: Test failed: h264dec: got 152 iterations transform.c:5023: Test failed: h264dec: got h264_encoded_data_len 1216 and no others. I'm not sure that I saw those at the time, but those at least can't be blamed to this patch. However, the added "got 53% diff" failures do indeed come from this patch series. Patch 2/8 is broken and I'm not sure why it works for you, since if I understand the comment correctly it still cannot work until we commit !5988. Thanks for looking into that. They still pass on my machine. I'll try to find out how they pass. To be clear, are there no issues with the implementation, only with tests needing todo_wine or perhaps flaky_wine?
-- https://gitlab.winehq.org/wine/wine/-/merge_requests/8812#note_129530