January 2023 - Wine-gitlab - list.winehq.org

[PATCH 0/1] MR1927: tools/winegcc: Reduce race conditions.
by eric pouech (＠epo) 13 Jan '23

13 Jan '23

When doing a parallel build on multi-arch tree, it sometimes generates errors like: /usr/lib/gcc/x86_64-w64-mingw32/12.2.1/../../../../x86_64-w64-mingw32/bin/ld: cannot find authz.dll-63db5999.spec.o: Aucun fichier ou dossier de ce type I added a couple of traces, showing that the two winegcc invocations for linking the DLL on the two arch:s are run in parallel *BUT* end up with the same output file. The one generated from tools.h / make_temp_file() called in winegcc{build_spec_obj()}. build_spec_obj() invokes 'winebuild -o XXX.spec.o' that starts (in -o option handling) by unlinking XXXX.spec.o. This opens a race condition on the output file name: 1) the XXX.spec.o is chosen by first arch in winegcc 2) XXX.spec.o is unlinked in winebuild (for first arch) 3) XXX.spec.o is chosen by second arch in winegcc 4) both winebuild invocations (first & second arch) now write to the same file In a nutshell, make_temp_file() isn't sufficient to ensure unicity of filename when such filename is erased (even briefly) in the build process. Proposed workaround: - use target name in generation of .spec.o filename to avoid collision. There's might other areas which could suffer from same kind of issue. Alternative options: - explicitly insert getpid() or target_alias in generated name in make_temp_file() - change value in make_temp_file (using time(NULL) + getpid() in parallel build can generate rather likely the same value in two consecutive processes) time(NULL) ^ getpid() might generate less collisions (even if it doesn't eradicate them). - change winebuild not to unlink output file but rather use open(O_TRUNC) (didn't look into it) Signed-off-by: Eric Pouech <eric.pouech(a)gmail.com> -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1927

2 3

[PATCH 0/1] MR1952: rpcrt4/tests: Avoid invalid free of static string.
by Rémi Bernon 13 Jan '23

13 Jan '23

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/1952

2 2

[PATCH v4 0/4] MR1683: loader: Support GDB hooking, and expose ld.so and PE modules in a wine-specific link map.
by Rémi Bernon 13 Jan '23

13 Jan '23

This makes it possible to debug Wine application in GDB almost like any other Unix application. Split from https://gitlab.winehq.org/wine/wine/-/merge_requests/1074. -- v4: ntdll: Maintain a PE module link map and expose it to GDB. ntdll: Pass a UNICODE_STRING to load_builtin and virtual_map_image. https://gitlab.winehq.org/wine/wine/-/merge_requests/1683

2 7

Re: [PATCH v2 0/4] MR1954: schedsvc: Avoid mismatching heap in several places.
by Rémi Bernon 13 Jan '23

13 Jan '23

v2: Fix `wine/heap.h` removal broken after some commit re-ordering. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1954#note_20848

1 0

[PATCH 0/4] MR1954: schedsvc: Avoid mismatching heap in several places.
by Rémi Bernon 13 Jan '23

13 Jan '23

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/1954

2 6

[PATCH 0/1] MR1953: comctl32: Avoid mismatching heap when in ProgressWindowProc.
by Rémi Bernon 13 Jan '23

13 Jan '23

-- https://gitlab.winehq.org/wine/wine/-/merge_requests/1953

1 1

[PATCH 0/2] MR1949: Draft: win32u: Correct GetRawInputBuffer alignment under WoW64.
by Tim Clem (＠tclem) 13 Jan '23

13 Jan '23

The existing mitigation for this (memcpy/overhead) leaves the first 8 bytes of RAWINPUT.data in the location of the wParam, and the upper 4 bytes of the wParam in the lower 4 bytes of the hDevice. Marking this as a draft due to code freeze. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1949

4 5

[PATCH 0/1] MR1950: widl: Initialize __frame to a default value (Coverity)
by Alistair Leslie-Hughes (＠alesliehughes) 13 Jan '23

13 Jan '23

In the generate code, Converity is reporting __frame->code unitilized in the failed case. When we initialize the object, then __f will also point to a fully initizliaed structure. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1950

4 6

[PATCH v2 0/2] MR60: vkd3d-shader: Merge all shader fork and join phases into a single phase.
by Conor McCarthy (＠cmccarthy) 13 Jan '23

13 Jan '23

SPIR-V doesn't support parallelisation in patch constant functions. Merging into one phase allows generation of a single patch constant function and normalises to the shader model 6 pattern. -- v2: vkd3d-shader: Merge all shader fork and join phases into a single phase. https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/60

2 2

[PATCH 0/6] MR61: vkd3d-shader/hlsl: Add field-level documentation to the most important structs in hlsl.h. (PART 2/2)
by Francisco Casas (＠fcasas) 13 Jan '23

13 Jan '23

Second part of !50. -- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/61

3 10