- 24 May, 2023 26 commits
-
-
Alexandros Frantzis authored
Store all incoming Wayland output state as pending and make it current in a thread-safe way when the "done" event arrives. This enables other threads, with proper locking, to safely read consistent and complete Wayland output information.
-
Alexandros Frantzis authored
Handle Wayland output events sent after process initialization, and update the win32u display devices when handling these events in the desktop window process.
-
Alexandros Frantzis authored
Create a dedicated thread from which we read and dispatch Wayland events (beyond the initial ones).
-
Alexandros Frantzis authored
There is currently no benefit to dynamic allocation, and static allocation allows us to simplify some aspects of initialization.
-
Zebediah Figura authored
-
Zebediah Figura authored
-
Zebediah Figura authored
-
Paul Gofman authored
VPN interfaces often have ARPHRD_NONE in ifr_hwaddr.sa_data but IFF_POINTOPOINT flag set.
-
Davide Beatrici authored
-
Davide Beatrici authored
-
Davide Beatrici authored
-
Anton Baskanov authored
-
Anton Baskanov authored
-
Anton Baskanov authored
-
Anton Baskanov authored
-
Anton Baskanov authored
It's less physically correct but closer to the native behavior.
-
Anton Baskanov authored
-
André Zwing authored
-
Alexandre Julliard authored
-
Alexandre Julliard authored
-
Alexandre Julliard authored
-
Alexandre Julliard authored
-
Alexandre Julliard authored
-
Alexandre Julliard authored
-
Alexandre Julliard authored
-
Alexandre Julliard authored
We are not going to support multi-machine .so modules.
-
- 23 May, 2023 14 commits
-
-
Davide Beatrici authored
programs/explorer/desktop.c:104:16: warning: ‘hres’ may be used uninitialized in this function [-Wmaybe-uninitialized] 104 | return hres; | ^~~~
-
Bartosz Kosiorek authored
Using TransformMatrixPoints is not needed and all values could be taken from transformation matrix: - ShearX from m11, m12 - ShearY from m21, m22 - Translation mdx, mdy The result could be calculated by taking destination points values: {{0.0, 0.0}, {1.0, 0.0}, {0.0, 1.0}} and calculating it with GdipTransformMatrixPoints function: dst_to_src_points[0].X = dst_to_src.matrix[4]; dst_to_src_points[0].Y = dst_to_src.matrix[5]; dst_to_src_points[1].X = dst_to_src.matrix[0] + dst_to_src.matrix[4]; dst_to_src_points[1].Y = dst_to_src.matrix[1] + dst_to_src.matrix[5]; dst_to_src_points[2].X = dst_to_src.matrix[2] + dst_to_src.matrix[4]; dst_to_src_points[2].Y = dst_to_src.matrix[3] + dst_to_src.matrix[5];
-
Bartosz Kosiorek authored
Improves efficiency by using addition instead of float numbers multiplication for destination points. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=53947
-
Bartosz Kosiorek authored
With previous implementation, every iteration pointer value was calculated by taking row and column of the image. It needs multiply calculation. With current implementation, pointer value calculation, is replaced with iterator, which takes next pixel data. It improves efficiency by using addition instead of multiplication for iterating through points data. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=53947
-
Bartosz Kosiorek authored
Calculating transformation points and not using it it is a waste of CPU time. With this patch the code was moved where it is used. It is improving performance of GdipDrawImagePointsRect when resampling is not needed. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=53947
-
Paul Gofman authored
-
Paul Gofman authored
-
Paul Gofman authored
-
Jactry Zeng authored
The default size is too small for HiDPI, so set the same height as the look-in combo box to it. And since in owner draw mode, the height of combo box is calculated and updated when drawing items, this patch set button height for toolbar from FILEDLG95_LOOKIN_DrawItem() too.
-
Jactry Zeng authored
-
Zebediah Figura authored
We test once above that the source does not use the existing allocator, but then implicitly release the allocator we are testing against (when the source assigns a new allocator to the sink). As a result, these ok() statements were testing that newly allocated allocators had a different address than the original allocator, which of course sometimes fails. Since we don't need to perform this test more than once anyway, just get rid of the offending ok() statements.
-
Ziqing Hui authored
-
Ziqing Hui authored
-
Ziqing Hui authored
-