On Android, backgrounding and foregrounding an app causes the Vulkan
surface to be destroyed. vkAcquireNextImageKHR returns
VK_ERROR_SURFACE_LOST_KHR, but the acquire while(true) loop only calls
RecreateSwapchain which doesn't recreate the surface, resulting in an
infinite retry loop and a black screen.
Handle VK_ERROR_SURFACE_LOST_KHR by setting both needsSurfaceRecreate
and needsSwapchainRecreate, then returning to let the existing
recreation path handle it on the next call.
Fixes#15322
This works inside of containers, and supports passing an activation token with the request, which is needed on Wayland to transfer focus to the browser.
(Strictly speaking, this was probably meant to be an "append" to the home/base
directory before the org/app name are appended to _that_, but it's definitely
added to the absolute start of the string on Emscripten, so might as well make
all of these match.
Reference PR #15262.
navigator.getGamepads() can return null for a slot if the gamepad
disconnects between the gamepadconnected event and the proxied
MAIN_THREAD_EM_ASM_INT call. This causes a TypeError when accessing
gamepad.id. Add null guards matching the pattern already used in
EMSCRIPTEN_JoystickOpen and EMSCRIPTEN_JoystickRumble.
The three EM_JS functions (SDL_GetEmscriptenJoystickVendor,
SDL_GetEmscriptenJoystickProduct, SDL_IsEmscriptenJoystickXInput)
call navigator.getGamepads() which is only available on the main
browser thread. With PROXY_TO_PTHREAD, the joystick callbacks are
dispatched to a worker where the Gamepad API is not available,
causing a TypeError.
Convert these from EM_JS to static functions using
MAIN_THREAD_EM_ASM_INT, which proxies the JavaScript execution to
the main browser thread. This matches the pattern already used by
other navigator.getGamepads() calls in the same file.
The rest of the atomic codepath is still enabled and usable.
This fixes missing and weird mouse cursors. We'll debug this code later on.
Reference Issue #15242.
Apparently this _can_ happen under load, or maybe some other weird condition.
Hopefully this will encourage PipeWire to fire output_callback again, and
we'll just try again later.
Reference Issue #14916.
- Initialize some out variables that are annotated inout in the function setting them.
- Fix 'dwVerHandle' might not be '0' warning from msvc analyzer calling GetFileVersionInfoA. MSDN says that
the lpdwHandle param to GetFileVersionInfoSize is optional (set to zero) and the dwHandle param to
GetFileVersionInfoA is ignored. msvc goes a step further and explicitly warns if dwHandle is not provably 0.
Fixes the following:
SDL3\src\stdlib\SDL_string.c(2359): warning C6054: String 'text' might not be zero-terminated.
SDL3\src\video\windows\SDL_windowsevents.c(897): warning C6001: Using uninitialized memory 'devName'.
SDL3\src\video\windows\SDL_windowskeyboard.c(644): warning C6388: 'dwVerHandle' might not be '0': this does not adhere to the specification for the function 'GetFileVersionInfoA'.
If the modern implementation of file dialogs on Windows fails, it will invoke the callback on error and report an error, then SDL would attempt invoking the dialog again using an older implementation, which would call the callback again.
This is not desired behavior, so it has been changed as follows:
- If the modern dialog fails before showing (missing library/symbol), don't call the callback and try the older dialogs in case we're on an old platform.
- If the dialog fails while or after showing, assume the error is a normal invocation error (such as an invalid path), call the callback and don't show the older dialogs to prevent showing two dialogs in succession to the user.
Visual Studio _still_ doesn't report itself as C99 compatible, afaict, but
does support the syntax as of VS2017 15.6, apparently.
This page mentions the first version of Visual Studio that handles hexidecimal
float notation:
https://stackoverflow.com/questions/18180116/vc-rejecting-hexadecimal-floating-point-constant?utm_source=chatgpt.com
If not Visual Studio, we also take the messier path for things that don't
report themselves as C99. Most things will take the cleaner path, though.
Closes#15276.
Caps the sample rate at 200kHz so that SDL's mixer will downsample any streams which are higher than that.
My Mega Drive emulator outputs at 223721Hz (the sample rate of the PSG chip), and `SDL_OpenAudioDeviceStream` fails due to DirectSound's `CreateSoundBuffer` returning an 'invalid parameter' error code. Lowering the sample rate makes the error go away.
Reported to me by @B3HKO in this issue:
https://github.com/Clownacy/clownmdemu-frontend/issues/60
The 200kHz limit is documented here:
https://learn.microsoft.com/en-us/previous-versions/windows/desktop/ee419022(v=vs.85)
DirectSound's Wikipedia article mentions that problems may occur with sample rates above 192kHz, but no source is provided, so I am unsure whether to take it seriously.
Now apps can have persistent files available during SDL_main()/SDL_AppInit()
and don't have to mess with Emscripten-specific code to prepare the filesystem
for use.
Fixes a crash when calling SDL_RenderGeometryRaw() with both `texture`
and `uv` set to `NULL`, and with geometry that is laid out in a way
that passes the quad checks.
with this, applications can do something like this:
SDL_SetHint(SDL_HINT_AUDIO_DEVICE_STREAM_NAME, "current song")
and this will be reflected in the Audio Devices Plasmoid, pavucontrol etc
also cleaned up PW_KEY_MEDIA_NAME and PW_KEY_NODE_NAME; setting them to the
same appears string appears to break convention, as it results in redundant
titles in PipeWire application lists.
The thread is still created once and only shut down on WIN_VideoQuit,
but the thread will unregister the devices when they're disabled to avoid
wasting time processing raw input that won't be appreciated by anyone.
This also allows RIDEV_NOLEGACY to be set on the devices, otherwise they
swallow up the WM_MOUSE* events that would be going to the main thread
when relative mode is disabled. That's out of scope for this change, though.
