So the game list watcher could rebuild the content providers
(CreateFactories) while the populate worker was still scanning them, and
the worker would walk a torn-down RegisteredCache and segfault in
OpenFileOrDirectoryConcat. Fixed by stopping and joining the worker
before rebuilding.
On macOS the watcher also re-armed itself every populate. Re-adding the
same paths makes QFileSystemWatcher re-emit directoryChanged (the
FSEvent comes in async, so the blockSignals guard misses it), so it just
kept refreshing and the list flashed forever. Now it only re-arms when
the watched dirs actually changed.
Also null-guarded OpenFileOrDirectoryConcat so a torn-down cache cant
null-deref there.
Related: https://github.com/eden-emulator/Issue-Reports/issues/336
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4099
Reviewed-by: Lizzie <[email protected]>
Reviewed-by: MaranBr <[email protected]>
The GL shader profile hardcoded max_user_clip_distances to 8. Query the
device limit like the Vulkan path already does (it reads
maxClipDistances), so we use what the host actually reports.
Clamp to Maxwell's NumClipDistances (8) since the guest never produces
more than that and the SPIR-V output array is sized for at most 8. So a
host reporting fewer than 8 is respected, and one reporting more can't
overrun anything.
Fixes#3910https://git.eden-emu.dev/eden-emu/eden/issues/3910
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4095
Reviewed-by: MaranBr <[email protected]>
Reviewed-by: Lizzie <[email protected]>
So CreateInstance was enumerating instance extensions twice, and that could race if the driver returned a different list the second time. On some AMD iGPU drivers, we could enable an extension from the first list, then fail the second check with VK_ERROR_EXTENSION_NOT_PRESENT.
Fix this by enumerating once and passing that same snapshot into RequiredExtensions.
Fixes: https://github.com/eden-emulator/Issue-Reports/issues/414
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4072
Reviewed-by: Lizzie <[email protected]>
Reviewed-by: MaranBr <[email protected]>
OpenSaveDataFileSystem was missing a few SaveDataSpaceId mappings, so it would hit ASSERT(false). This broke games that open cache storage, with TOTK being the obvious one, because save enumeration could assert before the game even finished loading.
Fixed this by mapping Temporary to user nand, and ProperSystem/SafeMode to system nand. These only get used for the free-space check here, so they just need to point somewhere sane.
Fixes: https://github.com/eden-emulator/Issue-Reports/issues/368
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4069
Reviewed-by: Lizzie <[email protected]>
Reviewed-by: MaranBr <[email protected]>
GetIndexFromAruid returns AruidIndexMax as the “not found” value, but both
GetControllerFromNpadIdType overloads were using that value to index
controller_data without checking it first.
when a game routes through DisconnectNpad before its applet resource is
registered, we can end up indexing past the end of controller_data. That seems
to be what was crashing some games on launch, especially on macOS and Android,
where the out-of-bounds read is less likely to just slide by unnoticed.
So yeah, AruidIndexMax is a sentinel, not a real controller_data index, so we
should not treat it like one.
Fixes: https://github.com/eden-emulator/Issue-Reports/issues/438
Fixes: https://github.com/eden-emulator/Issue-Reports/issues/439
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/4064
Reviewed-by: Lizzie <[email protected]>
Reviewed-by: MaranBr <[email protected]>
BoiledElectricity