Before, MetalFence was implemented as simply a busy loop on an atomic
int on metal, meaning the cpu would busy wait on the gpu to finish
taking power from it and decreasing battery life. This was the only kind
of cpu-gpu syncing (apart from requesting a swapchain)
(cherry picked from commit 514b26e4c4)
If a device has disconnected, and the app has quit the audio subsystem before
the queued main thread function has run, the pointer will be bogus by the time
the run occurs. So now it looks up the object from its device ID, and if the
lookup returns NULL, it returns immediately and everything works out.
Reference Issue #14856.
(which might be fixed by this or not.)
(cherry picked from commit 774728b6ff)
`Wayland_CreateAnimatedCursor` calls `SDL_qsort` on an array of
type `**SDL_Surface`, not on an array of `*SDL_Surface`. As such the void
pointers in the callback need to be casted to `SDL_Surface **` not
`SDL_Surface *`.
This was likely not caught since it's very unlikely that reading a few
bytes past the end of this array results in reading unreadable memory,
so the only side effect was invalid sort results, which is a bit subtle.
I caught this because I build SDL with UBSAN enabled in my debug builds,
and it trapped when the multiplication of the garbage `SDL_Surface*`
width and heights overflowed. I validated that the change is correct by
adding logs (that have then since been removed) demonstrating that it
was previously comparing garbage, and is now comparing the actual cursor
surfaces.
(cherry picked from commit a168e96cbc)
The event coordinates returned by GetMessagePos() for WM_NCACTIVATE are out of date if the cursor moved while an overlay was active, and may indicate that the cursor is still in the window when it is not. Always use the current cursor coordinates when processing this message to avoid incorrectly setting mouse focus if the cursor is no longer within the window.
(cherry picked from commit 0a8ccb11eb)
Add AndGAMER/Void Gaming USB IDs for Void GENESIS.
Recognize the Void GENESIS SInput VID/PID pair and set its device name.
Add DirectInput gamepad mappings for the wired and dongle DirectInput modes.
(cherry picked from commit 111bb79bf8)
libdecor is required for window decorations on Wayland when the toplevel decoration protocol is not supported, such as on GNOME and Weston. Warn if the development library can't be found, unless it was explicitly disabled.
(cherry picked from commit 607eb5206e)
The HIDAPI backend table used function pointers that didn't match the
macro-renamed backend function types, which trips Clang's function sanitizer.
Adding small wrappers so the backend table has the correct function types and
the cast to the backend-specific device type happens inside a direct call.
fixes#15821.
(cherry picked from commit 46d0ecead0)
Always pass the EGL platform type for Wayland and X11, or the driver could potentially select the wrong backend if certain envvars are misconfigured.
(cherry picked from commit 34af24276f)
A number of third party Xbox controllers are not supported by macOS, but work with libusb and the SDL HIDAPI driver.
(cherry picked from commit 18fc4d931a)
Otherwise, a device that is disconnected in the standard audio device thread
might keep failing WaitDevice() in a tight loop, each one generating a new
main thread callback. In normal situations, this is wasteful, but if the
app isn't pumping the event loop quickly (or at all!), this will quickly eat
up all the memory in a machine.
Now we note that the device is zombified right away, and device thread
iteration will use this to replace the implementation with the Zombie
equivalents once it owns the device lock.
The main thread callback will progress to device->zombie==2, which it uses to
decide if this is a duplicate disconnect notification. Since it also owns the
lock at this point, it takes the moment to set the Zombie implementation up,
too.
This allows things (like the WASAPI backend) to check for a non-zero zombie
state immediately without having to worry if the main thread callback ran, and
for the standard audio threads to also move to the Zombie implementation
without waiting on that callback.
(The Zombie implementation is used to make a dead device keep processing, so
things that need the audio device to make progress to function will keep
working, and things blindly pushing to an audio stream won't queue up endless
data that isn't being consumed.)
Fixes#15745.
(cherry picked from commit 6136358840)
