The `-Wstringop-overread` fix feels a bit questionable (so I'll add Eric as a
reviewer to clarify the MSC structs details)
--
v4: winedump: Initialize size variable in dump_dir_exceptions().
winedump: Use offsetof() for string position calculations.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5625
That seems to fix some (difficult to reproduce) crashes in EA Desktop / CEF, usually on shutdown but sometimes during starting up.
Currently TpSetWait can set (or clear) the event, while waitqueue_thread_proc() gets woken from NtWaitForMultipleObjects by previously set wait object and call the callback as if new set (or cleared) wait object is signaled. The crashes I was reproducing were always happening when RtlDeregisterWaitEx call is racing with waking the wait and calling the callback.
--
v2: ntdll: Make sure wakeups from already unset events are ignored in waitqueue_thread_proc().
https://gitlab.winehq.org/wine/wine/-/merge_requests/5044
This also bumps the minimum supported version of macOS to 10.12 to unconditionally assume the ulock syscalls are present.
As discussed with @Gcenx this shouldn't be an issue, as the current minimum support version of 10.8 does not work anyways at the moment with only 10.10+ being useable and even then virtually no-one using wine on versions < 10.12 (and also being unsupported by the prebuilt binaries).
Bumping to 10.12 would also have the side-benefit of allowing some cleanup of deprecated functions or legacy codepaths (like `mach_continuous_time` vs `mach_absolute_time` and probably much more).
In my testing ulock futexes are about 2x faster for synchronization than kevents. Using the syscall wrappers here should provide protection against changing syscall numbers, and given that ulock is also directly used by libc++, the probability of an incompatible API change here is basically zero imo.
By default ulock is also process-private, similar to the current linux implementation, cross-process synchronization would require darwin 19+ (macOS 10.15+) with `UL_COMPARE_AND_WAIT_SHARED`.
The diff is a bit bigger than it actually is in effect, since https://gitlab.winehq.org/mzent/wine/-/commit/28a8a991c165f0d37dbf78ca3cdee… is only moving some code around and not actually changing anything.
--
v4: ntdll: Implement futex_wait() and futex_wake_one() on macOS.
ntdll: Simplify futex interface from futex_wake() to futex_wake_one().
ntdll: Use USE_FUTEX to indicate futex support.
ntdll: Prefer futex for thread-ID alerts over kqueue.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5237
This is the first batch of a series implementing faster media source resolution required to workaround an Unreal Engine race condition present in some games, and deterministic stream ordering that decodebin / parsebin cannot provide, which is required to expose the streams in native order, for compatibility in several other applications.
I pushed the full series as a branch here: https://gitlab.winehq.org/rbernon/wine/-/commits/mr/wg-source-part-one
Note that this full series is also a first step in the direction of having a simpler demuxer interface, which will be required in the future for compatibility with applications that build MF or DirectShow pipelines directly and expect the relevant components to behave as a demuxer and expose compressed media types. For now it only delays the use of wg_parser to whenever the media source is started, and matches the non-ordered streams using their media types and tags. This is a best effort solution but I don't think we can do much better for the moment.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/3606
--
v2: winegstreamer: Expose the generic video decoder transform.
winegstreamer: Introduce a generic audio decoder transform.
winegstreamer: Rename aac_decoder to audio_decoder.
winegstreamer: Translate generic unknown audio / video media types.
winegstreamer: Support generic audio / video unknown formats.
winegstreamer: Call gst_video_info_from_caps for all video formats.
winegstreamer: Call gst_audio_info_from_caps for all audio formats.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5138
Adds the registry key
HKEY_CURRENT_USER\\Software\\Wine\\Wayland Driver\\rawinput
witch allows mouse raw input. This makes it easier
to calculate the same sensitivity in different games,
use sensitivity calculators, and easily change values
when changing mouse DPI and do not depend on the compositor or OS.
For example, you want to set the sensitivity to half as much,
but sensitivity curves in libinput are more difficult
to calculate than mouse sensitivity in the games.
Implementation of ideas written in the comments: https://gitlab.winehq.org/wine/wine/-/merge_requests/4698
--
v6: winewayland.drv: Add mouse rawinput support
https://gitlab.winehq.org/wine/wine/-/merge_requests/5869
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=44863
The bug has been fixed already by moving ddraw4 vertex buffers into
system memory. Changing this value makes the game create a smaller
buffer, which makes the game fast on video memory buffers. I think we
should stay close to what Windows drivers report even though we
mitigated the original issue in a different way.
--
v2: ddraw: Set dwMaxVertexCount to 2048.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5687
--
v2: d3dx9: Add support for mipmap generation to D3DXCreateVolumeTextureFromFileInMemoryEx().
d3dx9: Use d3dx_image structure inside of D3DXCreateVolumeTextureFromFileInMemoryEx().
d3dx9: Cleanup texture value argument handling in D3DXCreateVolumeTextureFromFileInMemoryEx().
d3dx9: Refactor texture creation and cleanup in D3DXCreateVolumeTextureFromFileInMemoryEx().
d3dx9/tests: Add more tests for D3DXCreateVolumeTextureFromFileInMemoryEx().
d3dx9: Use shared code in D3DXLoadVolumeFromFileInMemory().
https://gitlab.winehq.org/wine/wine/-/merge_requests/5881
Fix black screen issues for Active Trader Pro and Assetto Corsa. This reverts commit 4124478b.
For Active Trader Pro, the hack ends up clearing a context that was previously drawn to to black,
possibly because multiple OpenGL contexts are involved according to Henri's investigation.
Assetto Corsa creates a D3D device for a window but doesn't actually do any rendering with it. After
ceefcca7, having an OpenGL context for a window clears its OpenGL surface to black after its view
becomes visible. The OpenGL surface is on top of other layers for the window so the window becomes
black after that. The blackness is originally from glDrawBuffer(GL_FRONT_AND_BACK), which presumably
prepares a zero-initialized OpenGL front draw buffer. This also suggests that 4124478b is no longer
necessary. I could delay the clearToBlackIfNeeded() call so it uses the behavior before ceefcca7,
but now that it seems the hack is no longer necessary and might cause more errors. CrossOver has
disabled the hack for a while and it seems fine. I think it's time to remove the hack.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5929
As mentioned in https://gitlab.winehq.org/wine/wine/-/merge_requests/5264, the next step for OpenGL in the Wayland driver is support for WGL_ARB_pixel_format. It seems possible, at least in theory, to move a large part of the logic involved in this extension outside the drivers for common use by all of them (or ones that want to opt-in).
The goal of this RFC MR is, through experimentation and discussion/feedback, evaluate:
1. Whether making such functionality available outside the drivers (likely in an opt-in manner to begin with) is a productive way forward, or the complexity and platform specific decisions favor the current per-driver approach.
2. If we think that (1) is a worthy goal, what's the best mechanism to achieve it.
Note that the focus of this MR is currently on being a proof-of-concept, rather providing ready-for-detailed-review code (although I have done my best to keep the code decent). This MR currently (roughly in commit order):
* Introduces a new wgl driver callback to allow drivers to provide to opengl32/unix a list of formats and many details about them.
* Uses the information in that list to implement wglGetPixelFormatAttrib*.
* Uses the information in that list plus format sorting rules from WineX11 (effectively GLX rules plus tweaks) to implement wglChoosePixelFormatARB.
* Implements the get_pixel_formats callback for the Wayland driver.
* Hacks the get_pixel_formats callback for WineX11, and to allow me to run some experiments to compare the output of native WineX11 and get_pixel_formats-WineX11. In the admittedly not too many games I tried the sort order is the same, so at least that's encouraging.
My thoughts and notes so far:
* Using this approach for wglGetPixelFormatAttrib* works well, and we can also implement wglDescribePixelFormat in this way.
* It's not at all clear what the "right" sorting rules are for wglChoosePixelFormatARB.
* WineX11 uses GLX + tweaks (e.g., changes depth sorting). This means that larger formats tend to be preferred, at least according to the GLX spec. For example, asking for a r5, g6, b5 in the attributes is supposed to give back rgb888 (or even higher if available) as the top format in the list. Interestingly, and to confuse things even more, I haven't been able to make GLX return non-888(8) configs at all to actually test this more, so perhaps that's what saves it here? However, eglChooseConfig works similarly and there I was able to verify this behavior (e.g., got a nice surprise 10-bit format when asking for 5551 :)).
* Winemac has its own custom logic.
* Mesa's WGL implementation uses a different approach, closer to Wine's normal (wgl)ChoosePixelFormat, where proximity to the target format is strongly rewarded (so it seems asking for r5g6b5 is much more likely to actually get you that).
* Of course, the "gold standard" here would be to try to infer and use the rules used by some windows driver.
Looking forward to thoughts/feedback!
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5388
On the client side, frame aperture may be included or omitted. This can be used to include or drop the frame padding, and the VideoProcessor MFT supports it.
On GStreamer side, we can only use the unpadded frame size in the caps, because this is how decoders work, and padding needs to be added as a property of the input/output video buffers. Then, frame size needs to be adjusted to be consistent between input and output, and any difference considered as extra padding to be added on one side or the other.
--
v3: winegstreamer: Respect video format padding for input buffers too.
winegstreamer: Exclude padding from wg_format video frame size.
winegstreamer: Use video info stride in buffer meta rather than videoflip.
winegstreamer: Use a new wg_video_buffer_pool class to add buffer meta.
winegstreamer: Update the output_format when stream format changes.
winegstreamer: Pass optional GstVideoInfo to read_transform_output_data.
winegstreamer: Split read_transform_output_data in two helpers.
winegstreamer: Introduce a new sample_needs_copy_buffer helper.
winegstreamer: Introduce a new sample_set_flags_from_buffer helper.
mf/tests: Add more video processor tests with aperture.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5055
This MR adds a default implementation for `wglChoosePixelFormatARB` (in opengl32.dll) which will be used automatically if the driver populates the wgl_pixel_format ARB fields. Thus winex11 automatically uses this, and this MR removes the driver internal implementation.
The second part of this MR adds `WGL_ARB_pixel_format` support for winewayland (we can move this to a separate MR, but I wanted to have another implementation to better assess the behavior).
The `wglChoosePixelFormatARB` behavior is loosely modeled after the the AMD driver as discussed in https://gitlab.winehq.org/wine/wine/-/merge_requests/5388. I tried to keep things simple to begin with, and the results I am getting in various apps/games are sensible, but please try this out so we can refine further.
--
v2: winewayland: Support WGL_ARB_pixel_format_float.
winewayland: Support WGL_ARB_pixel_format.
winex11: Remove driver wglChoosePixelFormatARB implementation.
opengl32: Add default implementation for wglChoosePixelFormatARB.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5933
I have a question for Julliard before doing additional work:
1. Why do only MR test jobs have the extra-quiet winetest option enabled?
I think it exposes test flakiness (which could be insightful in normal Wine jobs too)
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5952
This saves ~1500 LoC, and at the same time allows mapping arbitrary long codec data back and forth GstCaps, which will be required to support arbitrary compressed formats. We can then for instance expose unknown codecs as Wine-specific subtypes, expose generic decoders and pass the backend GstCaps serialized in the codec data.
It makes the code simpler by unifying all frontend mappings and only requiring mapping between GstCaps and a unified set of formats. Mapping between IMFMediaType and the AM_MEDIA_TYPE formats is done using the standard MF functions and are tested and validated in mfplat tests.
--
This merge request has too many patches to be relayed via email.
Please visit the URL below to see the contents of the merge request.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5615
This MR adds support for creating file mapping objects backed by large pages on Linux, by making the following changes:
## wineserver
* On Linux, `create_temp_file` will first attempt to use memfds as the backing fd. If it fails, it'll return to the current codepath, creating a temporary file in either the server or config directory.
* The created memfd will be sealed against writes, if the caller requesting the appropriate page protection flags.
* This removes the requirement that FDs be only created on filesystems/directories that aren't `noexec`.
* In the server method `create_mapping` , if large pages have been requested by the caller, hold that the calling thread's token holds `SeLockMemoryPrivilege` .
* Additionally, add `SeLockMemoryPrivilege` to the list of privileges enabled for the Administrator.
## `ntdll`
* Add `virtual_get_min_large_page_size` and its exported wrapper `wine_unix_get_min_large_page_size`.
* On Linux, the minimum page size is determined by going through `/sys/kernel/mm/hugepages`. If hugepage support was not detected, `STATUS_NOT_SUPPORTED` is returned instead. On other platforms, the older hard-coded value of 2\*1024\*1024 is returned instead.
* `NtCreateSection` will validate certain parameters if large pages are requested. Specifically, it will return STATUS_INVALID_PARAMETER if the requested mapping is not anonymous/unnamed, or the size is not a multiple of the minimum supported page size.
## `kernelbase`
* `GetLargePageMinimum` will use `wine_unix_get_min_large_page_size`.
## `kernel32/tests`
* Add new test test_large_page_file_mapping, which validates privilege enforcements and parameter validation while creating large pages backed file mapping obejcts. The tests are skipped if `GetLargePageMinimum` returns 0.
--
v34: psapi: Add tests for querying information for large pages.
ntdll: Use PAGEMAP_SCAN ioctl to implement get_working_set_ex, if available.
kernel32: Add tests for large pages support.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5769
This MR adds support for creating file mapping objects backed by large pages on Linux, by making the following changes:
## wineserver
* On Linux, `create_temp_file` will first attempt to use memfds as the backing fd. If it fails, it'll return to the current codepath, creating a temporary file in either the server or config directory.
* The created memfd will be sealed against writes, if the caller requesting the appropriate page protection flags.
* This removes the requirement that FDs be only created on filesystems/directories that aren't `noexec`.
* In the server method `create_mapping` , if large pages have been requested by the caller, hold that the calling thread's token holds `SeLockMemoryPrivilege` .
* Additionally, add `SeLockMemoryPrivilege` to the list of privileges enabled for the Administrator.
## `ntdll`
* Add `virtual_get_min_large_page_size` and its exported wrapper `wine_unix_get_min_large_page_size`.
* On Linux, the minimum page size is determined by going through `/sys/kernel/mm/hugepages`. If hugepage support was not detected, `STATUS_NOT_SUPPORTED` is returned instead. On other platforms, the older hard-coded value of 2\*1024\*1024 is returned instead.
* `NtCreateSection` will validate certain parameters if large pages are requested. Specifically, it will return STATUS_INVALID_PARAMETER if the requested mapping is not anonymous/unnamed, or the size is not a multiple of the minimum supported page size.
## `kernelbase`
* `GetLargePageMinimum` will use `wine_unix_get_min_large_page_size`.
## `kernel32/tests`
* Add new test test_large_page_file_mapping, which validates privilege enforcements and parameter validation while creating large pages backed file mapping obejcts. The tests are skipped if `GetLargePageMinimum` returns 0.
--
v33: psapi: Add tests for querying information for large pages.
ntdll: Use PAGEMAP_SCAN ioctl to implement get_working_set_ex, if available.
kernel32: Add tests for large pages support.
ntdll: Support allocating virtual memory, creating and mapping files backed by large/huge pages.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5769
This MR adds support for creating file mapping objects backed by large pages on Linux, by making the following changes:
## wineserver
* On Linux, `create_temp_file` will first attempt to use memfds as the backing fd. If it fails, it'll return to the current codepath, creating a temporary file in either the server or config directory.
* The created memfd will be sealed against writes, if the caller requesting the appropriate page protection flags.
* This removes the requirement that FDs be only created on filesystems/directories that aren't `noexec`.
* In the server method `create_mapping` , if large pages have been requested by the caller, hold that the calling thread's token holds `SeLockMemoryPrivilege` .
* Additionally, add `SeLockMemoryPrivilege` to the list of privileges enabled for the Administrator.
## `ntdll`
* Add `virtual_get_min_large_page_size` and its exported wrapper `wine_unix_get_min_large_page_size`.
* On Linux, the minimum page size is determined by going through `/sys/kernel/mm/hugepages`. If hugepage support was not detected, `STATUS_NOT_SUPPORTED` is returned instead. On other platforms, the older hard-coded value of 2\*1024\*1024 is returned instead.
* `NtCreateSection` will validate certain parameters if large pages are requested. Specifically, it will return STATUS_INVALID_PARAMETER if the requested mapping is not anonymous/unnamed, or the size is not a multiple of the minimum supported page size.
## `kernelbase`
* `GetLargePageMinimum` will use `wine_unix_get_min_large_page_size`.
## `kernel32/tests`
* Add new test test_large_page_file_mapping, which validates privilege enforcements and parameter validation while creating large pages backed file mapping obejcts. The tests are skipped if `GetLargePageMinimum` returns 0.
--
v32: psapi: Add tests for querying information for large pages.
ntdll: Use PAGEMAP_SCAN ioctl to implement get_working_set_ex, if available.
kernel32: Add tests for large pages support.
ntdll: Support allocating virtual memory, creating and mapping files backed by large/huge pages.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5769
without freetype, some calling return failed, so some value may not be initialized (it initialize by some failed calling and pass argument by point of it), it would cause random crash!
Signed-off-by: Fan WenJie <fanwj(a)mail.ustc.edu.cn>
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5957
Signed-off-by: Fan WenJie <fanwj(a)mail.ustc.edu.cn>
when get_char_dimensions failed, width not set, width may be undefined value and cause undefined behaviour
--
v7: win32u: fix random crash without libfreetype
https://gitlab.winehq.org/wine/wine/-/merge_requests/5854
On Sat Jun 29 20:07:51 2024 +0000, Loïc Rebmeister wrote:
> @infyquest I'm happy you tried to make a fix, I'm happy you have
> interest in wine.
> > I dont need you to teach me code, Please don't make it personal.
> This ain't personal, if someone want to make a fix, they need at least
> to understand the code they send, this is the same with everyone.
> I know you wanted to do good, but this doesn't change what this code
> effectively does.
I am contributing to wine for more than 10 years. I admit that I might be wrong in understanding the code. Please try to communicate and present in a decent way as it will rub the people in a wrong way.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5771#note_74757
@infyquest I'm happy you tried to make a fix, I'm happy you have interest in wine.
> I dont need you to teach me code, Please don't make it personal.
This ain't personal, if someone want to make a fix, they need at least to understand the code they send, this is the same with everyone.
I know you wanted to do good, but this doesn't change what this code effectively does.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5771#note_74756
Please upload the code to the bug and test it. This was an attempt to fix the bug based on observations. There can be issues, with the code.
I dont need you to teach me code, Please don't make it personal.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5771#note_74751
Your pull request effectively make the code worse, as `{ L"Windows.Foundation.UniversalApiContract", 10, },` means `10` or lower.
So `{ L"Windows.Foundation.UniversalApiContract", 7, },` was already included by `{ L"Windows.Foundation.UniversalApiContract", 10, },`
But due to how `is_api_contract_present` work, it only take the first element with a name into consideration.
This is a simplified pseudo code with comments of `is_api_contract_present` to show the loop stop at first name match:
```java
boolean is_api_contract_present(String name,int version) {
for (present_contract : present_contracts) // List though all entries
if (present_contract.name == name)) // Check is name is equal
return version <= present_contract.max_major; // Check if version is lower than max_major, and return
return false; // No match found, return that we didn't found anything
}
```
Note: the `return` keyword stop code execution, and return the value.
(I'm assuming you didn't understood the code you sent there, as I'm pretty sure if you understood that code before sending it you wouldn't have opened this MR)
Further more, `Windows.Foundation.UniversalApiContract` is a dotnet 5 feature. (Source: https://github.com/madewokherd/wine-mono/issues/151#issuecomment-1232824011)
Wine (and Wine-mono) currently only support up to dotnet 4 on a clean install.
If you want I can teach you code for free so you can actually read and understand the code before you are sending it to wine upstream.
In the meantime I'll see what I can do to try to get the issue fixed.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5771#note_74750
This MR adds support for creating file mapping objects backed by large pages on Linux, by making the following changes:
## wineserver
* On Linux, `create_temp_file` will first attempt to use memfds as the backing fd. If it fails, it'll return to the current codepath, creating a temporary file in either the server or config directory.
* The created memfd will be sealed against writes, if the caller requesting the appropriate page protection flags.
* This removes the requirement that FDs be only created on filesystems/directories that aren't `noexec`.
* In the server method `create_mapping` , if large pages have been requested by the caller, hold that the calling thread's token holds `SeLockMemoryPrivilege` .
* Additionally, add `SeLockMemoryPrivilege` to the list of privileges enabled for the Administrator.
## `ntdll`
* Add `virtual_get_min_large_page_size` and its exported wrapper `wine_unix_get_min_large_page_size`.
* On Linux, the minimum page size is determined by going through `/sys/kernel/mm/hugepages`. If hugepage support was not detected, `STATUS_NOT_SUPPORTED` is returned instead. On other platforms, the older hard-coded value of 2\*1024\*1024 is returned instead.
* `NtCreateSection` will validate certain parameters if large pages are requested. Specifically, it will return STATUS_INVALID_PARAMETER if the requested mapping is not anonymous/unnamed, or the size is not a multiple of the minimum supported page size.
## `kernelbase`
* `GetLargePageMinimum` will use `wine_unix_get_min_large_page_size`.
## `kernel32/tests`
* Add new test test_large_page_file_mapping, which validates privilege enforcements and parameter validation while creating large pages backed file mapping obejcts. The tests are skipped if `GetLargePageMinimum` returns 0.
--
v31: psapi: Add tests for querying information for large pages.
ntdll: Use PAGEMAP_SCAN ioctl to implement get_working_set_ex, if available.
kernel32: Add tests for large pages support.
ntdll: Support allocating virtual memory, creating and mapping files backed by large/huge pages.
server: Use memfd to back anonymous mappings on Linux.
kernelbase: Implement GetLargePageMinimum by returning the value of LargePageMinimum in _KUSER_SHARED_DATA.
server: Set LargePageMinimum in _KUSER_SHARED_DATA on Linux.
server: Require SeLockMemoryPrivilege to create large page mappings.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5769
This MR adds support for creating file mapping objects backed by large pages on Linux, by making the following changes:
## wineserver
* On Linux, `create_temp_file` will first attempt to use memfds as the backing fd. If it fails, it'll return to the current codepath, creating a temporary file in either the server or config directory.
* The created memfd will be sealed against writes, if the caller requesting the appropriate page protection flags.
* This removes the requirement that FDs be only created on filesystems/directories that aren't `noexec`.
* In the server method `create_mapping` , if large pages have been requested by the caller, hold that the calling thread's token holds `SeLockMemoryPrivilege` .
* Additionally, add `SeLockMemoryPrivilege` to the list of privileges enabled for the Administrator.
## `ntdll`
* Add `virtual_get_min_large_page_size` and its exported wrapper `wine_unix_get_min_large_page_size`.
* On Linux, the minimum page size is determined by going through `/sys/kernel/mm/hugepages`. If hugepage support was not detected, `STATUS_NOT_SUPPORTED` is returned instead. On other platforms, the older hard-coded value of 2\*1024\*1024 is returned instead.
* `NtCreateSection` will validate certain parameters if large pages are requested. Specifically, it will return STATUS_INVALID_PARAMETER if the requested mapping is not anonymous/unnamed, or the size is not a multiple of the minimum supported page size.
## `kernelbase`
* `GetLargePageMinimum` will use `wine_unix_get_min_large_page_size`.
## `kernel32/tests`
* Add new test test_large_page_file_mapping, which validates privilege enforcements and parameter validation while creating large pages backed file mapping obejcts. The tests are skipped if `GetLargePageMinimum` returns 0.
--
v30: psapi: Add tests for querying information for large pages.
ntdll: Use PAGEMAP_SCAN ioctl to implement get_working_set_ex, if available.
kernel32: Add tests for large pages support.
ntdll: Support allocating virtual memory, creating and mapping files backed by large/huge pages.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5769
This MR adds support for creating file mapping objects backed by large pages on Linux, by making the following changes:
## wineserver
* On Linux, `create_temp_file` will first attempt to use memfds as the backing fd. If it fails, it'll return to the current codepath, creating a temporary file in either the server or config directory.
* The created memfd will be sealed against writes, if the caller requesting the appropriate page protection flags.
* This removes the requirement that FDs be only created on filesystems/directories that aren't `noexec`.
* In the server method `create_mapping` , if large pages have been requested by the caller, hold that the calling thread's token holds `SeLockMemoryPrivilege` .
* Additionally, add `SeLockMemoryPrivilege` to the list of privileges enabled for the Administrator.
## `ntdll`
* Add `virtual_get_min_large_page_size` and its exported wrapper `wine_unix_get_min_large_page_size`.
* On Linux, the minimum page size is determined by going through `/sys/kernel/mm/hugepages`. If hugepage support was not detected, `STATUS_NOT_SUPPORTED` is returned instead. On other platforms, the older hard-coded value of 2\*1024\*1024 is returned instead.
* `NtCreateSection` will validate certain parameters if large pages are requested. Specifically, it will return STATUS_INVALID_PARAMETER if the requested mapping is not anonymous/unnamed, or the size is not a multiple of the minimum supported page size.
## `kernelbase`
* `GetLargePageMinimum` will use `wine_unix_get_min_large_page_size`.
## `kernel32/tests`
* Add new test test_large_page_file_mapping, which validates privilege enforcements and parameter validation while creating large pages backed file mapping obejcts. The tests are skipped if `GetLargePageMinimum` returns 0.
--
v29: psapi: Add tests for querying information for large pages.
ntdll: Use PAGEMAP_SCAN ioctl to implement get_working_set_ex, if available.
kernel32: Add tests for large pages support.
ntdll: Support allocating virtual memory, creating and mapping files backed by large/huge pages.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5769
--
v2: winex11: Remove now unnecessary window surface BITMAPINFO.
winewayland: Remove now unnecessary window surface BITMAPINFO.
wineandroid: Remove now unnecessary window surface BITMAPINFO.
winemac: Remove now unnecessary driver surface BITMAPINFO.
win32u: Get rid of the unnecessary offscreen window surface struct.
win32u: Introduce a new helper to get surface color info and bits.
win32u: Pass BITMAPINFO and color bits to window surface flush.
win32u: Introduce a new helper to update layered window surface attributes.
win32u: Move layered surface attributes to the window_surface struct.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5947
This part XII of cmd engine rewrite.
This serie implements proper support for && and || command chaining operators:
- implement proper precedence parsing
- implement semantics for the operators
- for this to work, we need to have expected success/failure from commands;
this is supported here for external commands (and calling into a label or
another BAT/CMD file), but support has to be added for all builtin
commands
- that's a big mess as behaviors is far from homogeneous among the builtins:
+ some builtin set coherent success/failure vs errorlevel information
+ some report success/failure, but only set errorlevel upon failure
(keeping errorlevel from previous commands untouched),
+ some even report success (even in case of failure) yet set errorlevel
in case of failure,
- so the next series will now concentrate on testing success/failure and
errorlevel for every builtin command (sigh)
- for now, to help the transition phase, we can distinguish if a builtin
properly reports success/failure; if a builtin with success/failure
ability is used in a chaining operator, we implement the intented
behavior of the chaining operator; otherwise,
we keep old behavior (ie always execute LHS and RHS whatever the
success/failure status).
This shall be removed when all builtin commands report success/failure.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5949
This should reflect what we actually support, while still being a superset
of what native drivers actually support.
Native driver support varies, but, as the comments document, native drivers
(at least as of windows 10) support only RGB -> RGB, YUV -> RGB, and BC -> RGB
conversions. The actual formats supported are a more limited subset than exposed
(despite that the video cards should be perfectly capable of supporting more
conversions) but erring on the side of overreporting seems safer.
--
v3: wined3d: Implement wined3d_check_device_format_conversion().
https://gitlab.winehq.org/wine/wine/-/merge_requests/5816
This should reflect what we actually support, while still being a superset
of what native drivers actually support.
Native driver support varies, but, as the comments document, native drivers
(at least as of windows 10) support only RGB -> RGB, YUV -> RGB, and BC -> RGB
conversions. The actual formats supported are a more limited subset than exposed
(despite that the video cards should be perfectly capable of supporting more
conversions) but erring on the side of overreporting seems safer.
--
v2: wined3d: Implement wined3d_check_device_format_conversion().
https://gitlab.winehq.org/wine/wine/-/merge_requests/5816
This MR adds a default implementation for `wglChoosePixelFormatARB` (in opengl32.dll) which will be used automatically if the driver populates the wgl_pixel_format ARB fields. Thus winex11 automatically uses this, and this MR removes the driver internal implementation.
The second part of this MR adds `WGL_ARB_pixel_format` support for winewayland (we can move this to a separate MR, but I wanted to have another implementation to better assess the behavior).
The `wglChoosePixelFormatARB` behavior is loosely modeled after the the AMD driver as discussed in https://gitlab.winehq.org/wine/wine/-/merge_requests/5388. I tried to keep things simple to begin with, and the results I am getting in various apps/games are sensible, but please try this out so we can refine further.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5933
This is part XI of cmd engine rewrite.
This serie moves the IF and FOR instruction into the new CMD_NODE structure.
I'm fully aware it's a large change (commit #1).
It touches simultaneously at several components:
- lexer is no longer trying to handle nested instructions; this is now
handled over to parser which will take care of it in a recursive
fashion
- moving the IF and FOR instructions into CMD_NODE; this impacts the
parser and the execution part
These three parts are so tightly linked that I can't figure out a simple
way to split first commit in smaller bits.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5942
--
v4: d3d11/tests: Test UpdateSubresource() for NV12 textures.
d3d11/tests: Test NV12 textures without render target.
d3d11/tests: Do not check pitches.
d3d11/tests: Check for NV12 texture support before testing them.
d3d11/tests: Check the result of compiling HLSL shaders.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5818
This is another step towards getting render states out of wined3d_state.
The idea here is similar to the "push constant buffers" used for d3d8/9 flat
constants, and reuses the same infrastructure. That is, we pack all of the
constants directly into a wined3d_buffer which is bound directly to the
wined3d_state as a normal constant buffer. For SPIR-V this buffer is placed in
the GPU, and the shader can simply access it directly, with no renderer-specific
changes needed other than to determine the buffer's location.
Currently this buffer is still stored on the CPU for GLSL, and the contents are
pulled back from the sysmem resource pointer, instead of being pulled from
state->render_states.
Eventually, at least for Vulkan, the idea will be to write a FFP replacement
shader which takes structured data (which, given the constraints of SPIR-V, is
broadly necessary; we cannot simply upload individual named uniforms as we can
with GLSL). In fact this will take the form of an HLSL shader compiled to Shader
Model 2 bytecode. This allows us to, once again, reuse pretty much all of the
existing infrastructure and avoid writing more backend-specific code, and has
the effect that this FFP replacement works not only with SPIR-V but also with
GLSL, allowing us to eventually remove the existing GLSL FFP pipeline, and
enabling me to develop the HLSL FFP pipeline mainly using the GL renderer [which
is useful since the Vulkan renderer is still missing a few other features].
The full branch is available at [1], and passes all tests when using the GL
renderer.
[1] https://gitlab.winehq.org/zfigura/wine/-/tree/himavant8
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5815
I'm currently, very unsuccessfully, tracking down a crash on the unix side that only happens on Gitlab, and although I'm still unable to reproduce it, it then would be useful to have gdb to attach on segfault when I will.
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5367
GStreamer uses _SC_NPROCESSORS_CONF to determine 'max-threads'. On the
Steam Deck, this is configured to be 16 (which is double its number
of logical cores).
_SC_NPROCESSORS_CONF also disregards a process's CPU affinity, thus it
can create more threads than is useful, which ultimately wastes memory
resources.
Using affinity to set 'max-threads' addresses both these problems.
--
v6: winegstreamer: Set MAX_THREADS to 4 for i386.
winegstreamer: Use thread_count to determine 'max-threads' value.
winegstreamer: Use process affinity to calculate thread_count.
https://gitlab.winehq.org/wine/wine/-/merge_requests/5923
Rémi Bernon (@rbernon) commented about dlls/winegstreamer/unixlib.c:
> return STATUS_UNSUCCESSFUL;
> }
>
> + if (SUCCEEDED(NtQueryInformationProcess( GetCurrentProcess(),
> + ProcessAffinityMask, &process_mask, sizeof(process_mask), NULL )))
> + thread_count = popcount(process_mask);
> + else
> + thread_count = 0;
```suggestion:-4+0
if (!NtQueryInformationProcess(GetCurrentProcess(),
ProcessAffinityMask, &process_mask, sizeof(process_mask), NULL)))
thread_count = popcount(process_mask);
else
thread_count = 0;
```
SUCCEEDED is for HRESULT, Nt API returns NTSTATUS and we often just check that it's 0. Also note the style fixes (no space in paren here, continuation indent is 8 spaces).
--
https://gitlab.winehq.org/wine/wine/-/merge_requests/5923#note_74517