22 Jan
2025
22 Jan
'25
5:50 p.m.
On Wed Jan 22 17:50:27 2025 +0000, William Horvath wrote:
The instruction analysis isn't required in order for this check to fix the tests and the games. It might as well be vestigial in this patch, actually - it's evolved from what I initially thought the problem was, and what I thought was required for a correct solution. If it was simplified to just checking for a zeroed gsbase and replacing it with the Unix-side NtCurrentTeb, I don't see how that would introduce infinite loops that wouldn't have already been problems some other way?
That was also a bit discussed here FWIW and has some ad-hoc diff which fixes those tests (and additional tests attached there): https://gitlab.winehq.org/wine/wine/-/merge_requests/5480#note_67898