- Wine-gitlab - list.winehq.org

[PATCH 0/1] MR1390: winevulkan: Generate "in_ext" definition on demand
by Alexandros Frantzis (＠afrantzis) 16 Nov '22

16 Nov '22

There are cases in StructConversionFunction.definition where we will generate copy code for extension struct members, without emitting the definition of the "in_ext" variable used in the copy code. This issue is triggered by mismatches in the condition that guards the generation of the "in_ext" definition, and the condition(s) that govern the generation of the member copy code (e.g., in StructConversionFunction.member_needs_copy and VkMember.needs_conversion). In order to avoid such mismatches and the burden of having to keep the conditions in sync, this commit generates the definition on demand, by checking if it's actually needed by the member copy code. Signed-off-by: Alexandros Frantzis <alexandros.frantzis(a)collabora.com> -- Some more details in case a different solution becomes apparent. An example of the aforementioned issue can be triggered if we make VK_EXT_image_drm_format_modifier an UNEXPOSED extension in `make_vulkan`: ``` ../wine/dlls/winevulkan/vulkan_thunks.c: In function ‘convert_VkFormatProperties2_win32_to_host’: ../wine/dlls/winevulkan/vulkan_thunks.c:10214:119: error: ‘in_ext’ undeclared (first use in this function); did you mean ‘index’? 10214 | out_ext->pDrmFormatModifierProperties = convert_VkDrmFormatModifierPropertiesEXT_array_win32_to_host(ctx, in_ext->pDrmFormatModifierProperties, in_ext->drmFormatModifierCount); | ^~~~~~ | index ``` In this particular case the mismatch happens because for the `VkDrmFormatModifierPropertiesListEXT` (returnedonly=true) struct in the INPUT direction the current check doesn't emit `in_ext`: ``` if self.direction == Direction.OUTPUT or not ext.returnedonly: # Not triggered body += ident + "{0} *in_ext = ({0} *)in_header;\n".format(in_type) ``` but `self.member_needs_copy(ext, m)` below for the "pDrmFormatModifierProperties" member returns `True`, requiring `in_ext`. The flow is: ``` * self.member_needs_copy(ext, m) => True if m.name != "sType" and struct.returnedonly and not m.needs_conversion(self.conv, self.unwrap, Direction.INPUT, self.const): # Not triggered return False return True * m.needs_conversion(self.conv, self.unwrap, Direction.INPUT, self.const) => True # if pointer member needs output conversion, it also needs input conversion # to allocate the pointer if direction == Direction.INPUT and self.is_pointer() and \ self.needs_conversion(conv, unwrap, Direction.OUTPUT, struct_const): return True ``` -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1390

3 2

[PATCH v6 0/2] MR1334: xmllite/writer: Handle surrogate pairs
by David Kahurani (＠kahurani) 16 Nov '22

16 Nov '22

Surrogates have to be written in pairs. Also, handle related errors Signed-off-by: David Kahurani <k.kahurani(a)gmail.com> -- v6: xmllite/writer: Handle surrogate pairs https://gitlab.winehq.org/wine/wine/-/merge_requests/1334

2 3

[PATCH v7 0/4] MR741: ntdll: Implement NtFlushProcessWriteBuffers.
by Torge Matthies (＠tmatthies) 16 Nov '22

16 Nov '22

Based on [a patch](https://www.winehq.org/mailman3/hyperkitty/list/wine-devel(a)winehq.… by Jinoh Kang (@iamahuman) from February 2022. I removed the need for the event object and implemented fast paths for Linux. On macOS 10.14+ `thread_get_register_pointer_values` is used on every thread of the process. On Linux 4.14+ `membarrier(MEMBARRIER_CMD_GLOBAL_EXPEDITED, ...)` is used. On x86 Linux <= 4.13 `madvise(..., MADV_DONTNEED)` is used, which sends IPIs to all cores causing them to do a memory barrier. On non-x86 Linux <= 4.2 and on other platforms the fallback path using APCs is used. -- v7: ntdll: Add thread_get_register_pointer_values-based fast path for NtFlushProcessWriteBuffers. ntdll: Add sys_membarrier-based fast path to NtFlushProcessWriteBuffers. ntdll: Add MADV_DONTNEED-based fast path for NtFlushProcessWriteBuffers. ntdll: Implement NtFlushProcessWriteBuffers. https://gitlab.winehq.org/wine/wine/-/merge_requests/741

4 6

[PATCH 0/4] MR1373: mshtml: Page-wide events implementation.
by Gabriel Ivăncescu 16 Nov '22

16 Nov '22

This implements some events related to document/window load and unload / visibility changes. The latter is related in that native partially implements it, and only sends it when it is minimized or restored from being minimized, not on navigation or anything else. The main issue for me is the visibilitychange patch. Since we don't control the container hwnd, and it can even be in another process, there is no way to be notified cleanly when it changes its minimization state. We can't subclass it either, as the tests show, it's not subclassed. I implemented it now using an internal timer that polls for that, since it was the least invasive and works in all cases, including other process hwnd. Hooks wouldn't be able to do that, and are also more invasive IMO. I'm hoping for better ideas, though. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1373

3 7

Re: [PATCH v3 0/5] MR1365: msvcrt: add more *printf* implementations
by Piotr Caban (＠piotr) 16 Nov '22

16 Nov '22

Please don't update MR after it was approved (unless there's a reason to do that, like e.g. you have found a bug, there are conflicts during rebase). It invalidates the review. Since it was a simple rebase the patch still looks good for me. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1365#note_16031

1 0

[PATCH v2 0/2] MR1295: Enable compilation with long types in d3dx9_36/tests.
by eric pouech (＠epo) 16 Nov '22

16 Nov '22

-- v2: dlls/d3dx9_36/tests: Use correct integral type. d3dx9_36/tests: Introduce union to help passing float values in DWORD. https://gitlab.winehq.org/wine/wine/-/merge_requests/1295

3 4

[PATCH v2 0/2] MR1374: Create forward in kernel32 of MapViewOfFileFromApp and CreateFileMappingFromApp to kernelbase
by Santino Mazza (＠tati1454) 16 Nov '22

16 Nov '22

Some programs try to import these functions from kernel32.dll instead of kernelbase.dll -- v2: kernel32: Forward MapViewOfFileFromApp kernel32 to kernelbase. kernel32: Forward CreateFileMappingFromApp kernel32 to kernelbase. https://gitlab.winehq.org/wine/wine/-/merge_requests/1374

2 2

[PATCH v5 0/2] MR1334: xmllite/writer: Handle surrogate pairs
by David Kahurani (＠kahurani) 16 Nov '22

16 Nov '22

Surrogates have to be written in pairs. Also, handle related errors Signed-off-by: David Kahurani <k.kahurani(a)gmail.com> -- v5: xmllite/writer: Handle surrogate pairs xmllite/writer: Pass xmlwriteroutput directly. https://gitlab.winehq.org/wine/wine/-/merge_requests/1334

2 2

[PATCH 0/2] MR1374: Import MapViewOfFileFromApp and CreateFileMappingFromApp into kernel32
by Santino Mazza (＠tati1454) 16 Nov '22

16 Nov '22

Some programs try to import these functions from kernel32.dll instead of kernelbase.dll -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1374

4 7

[PATCH 0/1] MR1388: msv1_0: Drop the ntlm_auth version check.
by Hans Leidekker (＠hans) 16 Nov '22

16 Nov '22

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=53932 -- https://gitlab.winehq.org/wine/wine/-/merge_requests/1388

2 1