This fixes a bug when swapping NVIDIA effects.
The update signal triggers a destruction of the previous effect, but the
effect would still run, potentially causing a crash.
The mutex added in the NVIDIA processing function prevents that.
Signed-off-by: pkv <pkv@obsproject.com>
The code assumed MFXCreateSession would always succeed, but it fails on
systems with no QSV implementations, causing a crash when we call
MFXClose later. Additional success checks for the other API calls were
also added.
CEF apparently doesn't like it and is unable to handle when you
destroy/recreate CEF instances very quickly, so... let's just put a
timer on this insanely terrible function. I guess. Whatever.
If you're using an encoder from a plugin, and you remove that plugin,
this value will be null, causing a crash. (Lain note: I realize that
this isn't the best solution to this problem, though there really isn't
any way to infer what type of encoder the prior encoder was. Should also
save the encoder type as additional setting metadata in future
versions.)
Setting AdaptiveCQM to ON when LowPower is OFF crashes the
intel-oneapi-gpu implementation on Linux when using the AV1 encoder.
But using the same LowPower settings as Windows works so lets drop the
current guards.
We need to check for new versions of the NVIDIA Effects redistributables
because mismatched versions of the AUDIO and VIDEO sdk can have
different CUDA versions of the deps, which has caused crashes in obs.
It'd be way better that NVIDIA either ships a single installer with both
audio and video effects or that they auto-install with the drivers.
But meanwhile, we provide a warning to users which requires us to keep
in sync with what's released.
Signed-off-by: pkv <pkv@obsproject.com>
generate-docs action does not require a commitHash (it discovers this
value automatically) and redirects do not need to be set up anymore
as the old documentation pages have been removed.
When the internal git reference is nonexistent, GitHub Actions seems
to convert that value into a hash made up of only zeroes. As that
hash will be commit-ish it will pass verification.
git cat-file will fail if the provided reference cannot be resolved
in the repository so should be the safer choice.
Using older file system (HFS+ instead of APFS) and compression
(lzfse instead of lzma) allows the disk image to be opened on older
versions of macOS, avoiding the cryptic error message that users
are greeted with otherwise.
This is a temporary fix for the still-alive installed base of
Intel-based Macs.
VPL does not have a consistent initialization process for the various
backends it dispatches to. So this resolves the new VPL dispatcher
failing to initialize the video session when it dispatches to
intel-mediaSDK backend. On the oneVPL-intel-gpu backend VA-API sessions
are correctly initialized so this wasnt noticed when tested on newer
hardware.
In c8d95005c1, it was incorrect assumed that just because the fix
worked on OpenGL, that the fix would also work on D3D11. Automated tests
would be ideal here, although if there were automated tests, ironically,
it wouldn't have picked this up, because we (read: I) made D3D11 fall
back to OpenGL of D3D11 failed. Basically there is no one to blame but
myself again.
This time, we've removed the OpenGL fallback in a06d6893b9 (thank you
Rodney), and additionally I properly tested on both rendering backends.
(Lain note: I'm just going to go sit in the corner for a bit and
ruminate after this)
After commit 8955338624, 'UI/update/models/whatsnew.hpp' is included
unconditionally, even with 'ENABLE_WHATSNEW=OFF', which breaks
building OBS Studio without nlohmann JSON.
Due to my jank transpiler, default.effect worked with Direct3D 11 but
not with OpengGL. Fixing this particular bug with the transpiler is a
bit non-trivial so instead just modify the shader to work with both.
The use of GetCurrentSceneItem can lead to a race condition between
the graphics thread (where this function will be run) and the main UI
thread, as both access the scene list model and iterate through its
members which can change during iteration (as there is no lock around
its access).
When `obs_output_t` is an encoded output, `obs_output_set_media` will
ignore the video and audio so that `obs_output_t` will keep holding the
`video_t` and `audio_t` pointer when the output was created.
By this commit, `video` and `audio` member variables in `obs_output_t`
will never set if it is an encoded output.
In the auto-configuration wizard, `video_t` is released and created to
have a different video size while `obs_output_t` is not released. This
resulted in accessing the released `video_t` pointer.
Firstly, removes the UpdateTransformShortcuts() method introduced in
c33fa8b which was trying to re-implement the behavior of
UpdateEditMenu().
Secondly, updates UpdateEditMenu() to account for sources without video.
Sources without video shouldn't be able to have their transform edited,
copied, pasted, or changed in any other way (because they don't have
one).
Attempting to set the window title of a BrowserDock that is closed, then
opening the BrowserDock, would show that BrowserDock with an incorrect
title. We can handle this by overriding the showEvent of BrowserDock and
manually setting the window title after the showEvent is called,
hopefully ensuring that we are only setting the window title on a window
that exists.
Instead of relying on the dock's having their window title already set
for the menu items to be updated, let's just manually update the menu
item text ourselves when updating the item.
This is a workaround for a change in Qt behavior between 6.5.1 and
6.5.2[1] which affected the outcome of attempting to set the window
title of a window that is not currently open. In Qt 6.4.3 and 6.5.1, the
window title would be set. In Qt 6.5.2, the window title will become
NULL/empty.
Instead of relying on Qt to have valid titles for custom browser dock
windows that we manage, let's track the names ourselves.
[1]: c153066baa
Works around a bug in Qt where setWindowTitle called on a native window
that hasn't been initialised yet causes the title data to be lost,
resulting in the window's title never being set.
The resize event was triggered during the shtudown of OBS. A commit
9f330050ef moves the actual resize code in `gl_update` to the main
thread without synchronizing between the main thread and the graphics
thread. When the display context is destroyed just after the resizing,
the delayed resize code can read the destroyed display context.
As a workaround, this commit will destroy the display context at the
earlier timing of the shutdown process.
The display context was sometimes created again by a resize event after
destruction from `OBSBasic::closeEvent`. That resulted in the display
context alive until reaching the destructor of the `OBSQTDisplay` class
and caused a crash in `render_display` on macOS. To avoid this, destroy
the display context at the event `SurfaceAboutToBeDestroyed` and let not
create the display context again.
The commit e67fdbca79 is a workaround for QtWayland bug that the
`QEvent::Expose` was never received on a certain condition.
The bugfix was introduced to QtWayland 6.1.0 so that the workaround is
not necessary for now.