On Wed Aug 16 21:48:02 2023 +0000, Bartosz Kosiorek wrote:
> How it could be done (set \*result when you return)?
I just mean set *result before you return, just like you do about 5 lines before.
It appears 0 would the appropriate value, in which case you could combine the if blocks.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/3581#note_42603
Jeffrey Smith (@whydoubt) commented about dlls/gdiplus/tests/pathiterator.c:
> GpPath *path;
> GpPathIterator *iter;
> GpStatus stat;
> - INT start, end, result;
> - BOOL closed;
> + INT start, end, result, closed;
I would leave the type as BOOL, to make it match the 4th parameter of GdipPathIterNextSubpath.
But since it is implemented as an integer (as opposed to an actual boolean that only accepts true/false), it can contain any value that an INT can, so the tests should still be fine.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/3581#note_42601
I stumbled onto this while using freopen() for debugging purposes.
Basically, freopen() fails if the FILE has been created with
an invalid handle.
So, this MR contains:
- basic tests for freopen (no issue there, just for coverage purposes)
- tests for freopen on FILE with invalid handle
- fix for freopen
--
v3: msvcrt: Fix freopen() on FILE with invalid underlying fd.
msvcrt/tests: Add tests for freopen().
https://gitlab.winehq.org/wine/wine/-/merge_requests/3578
On Wed Aug 16 20:20:19 2023 +0000, Jacek Caban wrote:
> One way or another, tests of a global state like that will always be
> theoretically a subject to race. In essence, my patch reduces problem
> probability, just like your solution.
Which is why a general solution would be better than N different hacks.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/3577#note_42592