Rather than setting NetworkManager to "unmanage" the Wi-Fi interface every sync/connection, do it once when the pipe starts. This way, there won't be any awkward shuffling as the pipe goes from syncing to connecting.
* optimizations
* Update menu_game.h
* try new code
* use cached EGLImages
* try drm cache on pi
* nani?
* huh?
* require opengl egl
* find EGL
* find EGL separately
* logging
* add support for versions of ffmpeg less than 6.0
* disable drm cache for now
* Update ui_sdl.c
* use old code for now
* disable avcc hdr for now
* restore error message
* try disabling format modifiers?
* tell me what the layer is
* pi decoder actually produces frames in yuv420p
* print more debug info
* try this
* switch to faster decode path
* switch to old code
* Update ui_sdl.c
* add more timing profiling
* more printing
* enable some optimizations
* draw menus when not in game mode
* call render present
* Update ui_sdl.c
* Update ui_sdl.c
* remove unused code
* Update CMakeLists.txt
* incorporate direct DRM code
* Update CMakeLists.txt
* free drm context and test old decoding code
* move SDL event loop to separate thread so controls are not limited by display rate
* add a wait to reduce CPU thrashing
* fix double free
* separate video into recv and consume threads
* Update ui_sdl_drm.c
* Update ui_sdl_drm.c
* Update gamepad.c
* further optimizations
* anti-alias buttons more
* fix some compiler errors
* only include DRM on linux
* fix windows compile issues
* fix includes
* Update CMakeLists.txt
* add htobe32 define for macOS
* Update video.c
* Update gamepad.c
* initial commit
* seems to work
* try SDL_SetHintWithPriority
* set hint with SDL_HINT_OVERRIDE
* try setting hint before SDL_init
* fix issue with signal handling
* send sigterm to pipe
* ensure SDL creates signal handlers?
* fixed pipe issues not closing
* ensure pipe is stopped ASAP when receiving SIGINT and SIGTERM
The Steam Deck gives us extremely little time to clean up before getting SIGKILL'd, so make sure pipe is told to quit ASAP when receiving SIGINT or SIGTERM.
* fix logic issue
* removed things that aren't necessary for this PR
* clarified some comments [skip ci]
---------
Co-authored-by: MattKC <[email protected]>
Greatly improves routing behavior.
Rather than letting dhclient perform routing and then "correct" it after the fact, we replace dhclient's behavior entirely.
Rather than add a system-wide route to 192.168.1.10, we bind our socket to the interface so no conflicts will occur with any other device on the network named 192.168.1.10. No system-wide route is made via the wireless interface for 192.168.1.0/24 or for 192.168.1.10.
Fixes#96
itsmattkc