- Aug 28, 2023
-
-
Provides a clearer picture of what the shader is doing. Useful for debugging, and user feedback (since it shows e.g. what tone-mapping mode is active and what the source/target nits are).
-
-
Handily eliminates a bunch of needless repetition of this logic, as well as allowing us to support higher versions of GLSL in vlc_gl_sampler.
-
This allows vlc_gl_sampler to be used with the appropriate version of GLSL enabled, which allows the use of more up-to-date GLSL features. To avoid unnecessary complicated preprocessor interventions, we limit the upper version of desktop GLSL to 410, which is the last version that retains compatibility for attribute/varying declarations.
-
To be re-used.
-
-
This patch adds abstraction for vlc_Logger, via the Rust Logger type and similar macros as the C side with: debug!, info!, warn! and error!.
-
-
-
-
If we copy the chroma from glfmt_in, we have to copy the RGB mask as well otherwise they may not end up being matching values.
-
video_format_FixRgb() only sets the RGB mask if it's not set. So the call will have no effect. Furthermore the masks it would set for VLC_CODEC_RGB32 don't match the ones set locally. So it's probably not the call we want, unless the local masks are wrong.
-
Although with push we should not need this hack anymore. Especially if we assume we have to use the palette from each picture and not the format.
-
The chroma is set by the user, it may be one of the RGB with a mask chromas. And the user can't specify a mask, so later is will be assumed to be using these masks.
-
-
-
This will use the same masks as we did, but it's better to explicitly say we use the default masks.
-
-
-
As mentioned in previous commit, it's possible for the value parsing function to be called on a shorter string of a longer valid format, which should not be parsed in that case. The problem didn't arise previously because the function could only be called on null-terminated strings and those strings were only built from the split of ` --> ` into two strings.
-
Refactor the parsing of timing values so that it doesn't allocate memory. Indeed, we just need to parse the input string to locate where the delimiter is, and call the parsing of timing values on each side of the arrow. We also use a length validation defensively, despite the current impossibility to parse characters like `-->` in the sscanf given the current format, to ensure that future addition here would not try to parse additional characters and to ensure that future usage of the function on strings that were not splitted by such character before (eg. 0:0:0, 0 limited on 6 characters) don't impact the parsing.
-
The test is embedded into the subtitle.c file because it tests some internal functions parsing the input subtitle format. For now, only VLC_TICK_0 origin is tested, no specific example, but invalid strings are also tested.
-
Currently, 0 is mapped to VLC_TICK_INVALID. Since the timings will be reported as-is to the core afterwards, they need to start at VLC_TICK_0. It won't change the duration, given that i_start - i_stop removes the constant VLC_TICK_0 offset.
-
- Aug 27, 2023
-
-
If for some reason there's alpha data, it should also be found in the snapshot. Snapshots are stored as PNG by default, so it can handle RGBA sources.
-
The meson job will not use the contribs, and the contribs are already built separately in the debian-contrib job, so remove them from the build.
-
The debian job is not using the contribs at all, split the job in two so that it can be parallelized more easily and so that the debian job reflect the time it took to build VLC. Note that it notably changes the folder where debian contribs are uploaded.
-
This is my m4 syntax highlighting.
-
Either D3D11/D3D9 is usable and the structures are there, or they are not. We don't support old mingw-w64 versions without these headers. * D3D11 is not enabled if libavcodec/d3d11.h is not usable * DXVA2 is not enabled if libavcodec/dxva2.h is not usable * direct3d9_filters is enabled if d3d9.h is usable (ie not UWP builds)
-
Qt plugin libs depend on Qt libs, not the other way around. They should appear first when telling the linker.
-
functions from this library are used directly, but library is not directly linked. It only worked because Qt links dwmapi itself.
-
-
The function has a declaration in MinGW headers since v3.0.0, thus this patch should not be needed anymore.
-
This was fixes upstream by 44cdeeb70db9112be132a4fb5b21d0cb8d57acaf.
-
- Aug 26, 2023
-
-
Report the coverage from meson and list the output of the tests, since meson already supply a junit-compatible output.
-
Meson 1.0.0 allows writing 0x0 characters into the junit test output, which is not allowed in XML files and it prevents gitlab CI from parsing the output of the tests. This can be checked locally with: meson setup build meson test -C build css_parser xmllint build/meson-logs/testlog.junit.xml -
Add support for gcovr.
-
Missing from commit 4be29e38.
-
The NSMenu indexOf… methods return -1 and not NSNotFound, so the check here is insufficient and can lead to crashes due to passing a negative integer to itemAtIndex.
-
Use the official screensHaveSeparateSpaces API instead of manually reading User Defaults values, this is more reliable especially as changing this setting is not immediately applied only after logging out and back in, so previously we could get the wrong value as it would immediately update in the User Defaults even when not in effect yet.
-
- Aug 25, 2023
-
-
Steve Lhomme authored
If the source has alpha, we should keep it during the conversion to RGB.
-
