On Wed Jun 4 20:21:59 2025 +0000, Bernhard Übelacker wrote:
Hello @dlehman25, I noticed the test pattern page shows following tests starting to fail since 2025-05-28: - [d3dcompiler_42:hlsl_d3d9](https://test.winehq.org/data/patterns.html#d3dcompiler_42:hlsl_d3d9) - [d3dcompiler_43:hlsl_d3d9](https://test.winehq.org/data/patterns.html#d3dcompiler_43:hlsl_d3d9) - [d3dcompiler_46:hlsl_d3d9](https://test.winehq.org/data/patterns.html#d3dcompiler_46:hlsl_d3d9) - [d3dcompiler_47:hlsl_d3d9](https://test.winehq.org/data/patterns.html#d3dcompiler_47:hlsl_d3d9) A git bisect led me to this merge request containing this commit: ``` 15485e6dd9be152286f08b3a3f6cf42895f3b377 is the first broken commit commit 15485e6dd9be152286f08b3a3f6cf42895f3b377 Author: Daniel Lehman <dlehman25(a)gmail.com> Date: Fri Apr 25 21:45:19 2025 -0700 msvcr120: Add cexp() implementation. ``` (git bisect needed a clean build each time, so ccache helps a lot. Only visible with a GCC-mingw build.) I tried to further reduce it and think it is related with the addition of `-fno-builtin` to `libs/musl`, can you confirm this finding? As changing EXTRADEFS causes no immediate rebuild, I manually removed object files to find the object file, which causes test to fail. After having `libs/musl/src/math/__rem_pio2.c` rebuilt this way the test started to fail.
-- https://gitlab.winehq.org/wine/wine/-/merge_requests/7748#note_105439