Commit graph

84 commits

Author SHA1 Message Date
b3d54ff160 add doomlib which includes doomdef.h to fill out dp_doom_event_t structures
Those dp_doom_event_t structs are doom even_t structs...
2023-02-23 21:57:17 +01:00
6d03cf86f9 ctrl: fix bogus elapsed time calculation. doh! 2023-02-23 21:55:19 +01:00
663be5c54b prep for input publishing to dooms 2023-02-23 07:12:18 +01:00
8b9f391c53 dp_doom: force C11 for _Static_assert() 2023-02-23 06:17:38 +01:00
92bfba237f dp_doom: fix two warnings in old linuxdoom code 2023-02-23 06:16:43 +01:00
f1281295ee add cmake option to force using the included version of SDL2 2023-02-23 06:16:09 +01:00
9ef9e71a46 ctrl: add menu entries to show dear imgui userguide and about window 2023-02-22 22:35:51 +01:00
407dded717 cleanup and revive the debug drawing code 2023-02-22 21:23:31 +01:00
abb563440d add comment about DoomFrames from unregistered dooms 2023-02-22 21:15:22 +01:00
ff89a984fb set ui log level from trace to debug 2023-02-22 21:15:03 +01:00
390413c09d yes, it's buggy 2023-02-22 21:13:18 +01:00
9d097277cd copy wildmidi.cfg to the build directory so that music works by default 2023-02-22 20:31:00 +01:00
4efcc3cb95 look for doom1.wad and create a link in the build directory if found 2023-02-22 20:30:12 +01:00
348bcc9997 use the included SDL if not appropriate system-wide SDL was found 2023-02-22 20:29:41 +01:00
b6f81cf375 add SDL-2.26.3 2023-02-22 19:20:36 +01:00
8d141da463 Squashed 'external/SDL/' content from commit adf31f6ec
git-subtree-dir: external/SDL
git-subtree-split: adf31f6ec0be0f9ba562889398f71172c7941023
2023-02-22 19:20:36 +01:00
98c44f66df preparations for also using nng_tcp as a transport 2023-02-22 19:17:06 +01:00
8b5231a782 ctrl: reenable the waitpid() code to catch crashing/killed dooms
TODO: implement DoomState::tLastActive timeout logic
2023-02-21 23:05:53 +01:00
eea3377406 finally fix dooms not being able to output sound
It was just the empty environment after posix_spawn() that caused
SDL_INIT_AUDIO to fail. __environ fixes that.
2023-02-21 23:03:39 +01:00
0639476d91 cleanup ib_sound_nng.c There never was an issue there... 2023-02-21 23:01:31 +01:00
fb3441f5d3 toggle fullscreen, toggle ui, invert mouse panning 2023-02-20 22:50:17 +01:00
4a0117af36 fix crash due to logging during shutdown from class based DoomState
Crash was double free in ImGui::AppendLog() during shutdown. Will have to
disable the logging callback if logging should happen at that point.
2023-02-20 00:06:28 +01:00
e7c98a490e disable texture info debug 2023-02-20 00:01:29 +01:00
284e5d1b0e ctrl: fix rendering, erase Endooms, implement zooming and panning 2023-02-19 23:52:54 +01:00
cbe6f76920 dp_doom: use 4 bytes per pixel and fix the colors 2023-02-19 23:48:02 +01:00
3c5b3ae4e9 it's alive! not pretty but alive! also remove the fork() spawner 2023-02-19 16:33:23 +01:00
df9cf1c14a intermittent ctrl commit: fork() based spawn code which makes nng panic 2023-02-19 16:23:29 +01:00
fc302969f6 dp_doom: hacky way to run without audio
This needs fixing in the controller. Try something similar to xdg-open
where a doomgrandchild would execvp() the doom.
2023-02-19 16:22:51 +01:00
e53023079e ctrl: register externally started dooms that introduce themselves
TODO: Check on those via kill(pid, 0) or store the last activity time in
DoomState and let dooms time out if they don't send more updates.
2023-02-19 12:46:17 +01:00
9e857180be dp_doom: fix silly bug causing only the first frame to be published 2023-02-19 12:40:26 +01:00
c292aadb68 dp_doom: make errors from IB_StartupSound() non-fatal
Not sure if it actually runs without sound or die horribly. Needs testing.
2023-02-19 12:39:43 +01:00
c51f83a5fa dp_doom: switch to own sdl sound impl because issues
Wasn't a bug in the sound code at all. This version is just cleaned up
and is not compatible with SDL1 anymore.
2023-02-19 12:38:25 +01:00
cd25a3380a unique filenames for ib_{sound, system, video}/ implementation files 2023-02-19 10:54:31 +01:00
381b8bd03c silence doomsim logging even more 2023-02-19 10:46:26 +01:00
236652be35 make dp_doom send out framebuffers via nng 2023-02-19 10:46:12 +01:00
366bf9b75c ctrl: better networking, can render received frames, still a broken mess 2023-02-19 10:44:57 +01:00
0e7eb1f5b8 fill out some of the doom implementation 2023-02-19 03:08:56 +01:00
777d766035 rename clowndoom to dp_doom, add skeleton code, fix issues 2023-02-19 02:17:52 +01:00
c9c334ae57 add modified clowndoom 2023-02-19 01:52:56 +01:00
a191a82bdc Squashed 'src/dp_doom/' content from commit 1909924
git-subtree-dir: src/dp_doom
git-subtree-split: 190992421c7c643afc99d7f7c511c162f65bea85
2023-02-19 01:52:56 +01:00
dfbfb4b08c and back to spawning doomsims 2023-02-19 01:46:31 +01:00
773ea7341c intermittent commit: do not spawn doomsim instances to test rendering
Without spawning doomsims the SDL_UpdateTexture()+SDL_RenderCopy() way does
scale well: with 1024 non-spawned dooms I'm at 19.4 fps, 51.5 ms/frame.
A single core is maxed out all the time and it's nicely jumping between
cores.
2023-02-19 01:45:51 +01:00
feb0bc995c work on rendering - can draw rectangles into buffers and display those
Approach is to have one SDL_Texture per doom and use SDL_RenderCopy() with
different destination rectangles to compose the layout. Apart from
SDL_UpdateTexture() and SDL_RenderCopy() no work is done per doom.

This scales up to ~200 doomsim instances. After that the frametime
increases rapidly and all my cores are maxed out.

TODO: test without spawning doomsim processes!
2023-02-19 01:24:52 +01:00
9dab0ccf70 rename QuitDoom -> Endoom :) 2023-02-18 23:04:34 +01:00
e590bed17a slight render call rearrangement 2023-02-18 22:41:05 +01:00
86dcd6f92d get rid of the doomId concept, use the pid instead
The internal unique id should not be needed, the pid is unique enough :)

Also this can allow starting dooms externally without having to pass an id
parameter that's not already in use. The externally started doom sends its
pid with the DoomReady message and the controller can add a new entry if it
needs to. Of course the controller will not be able to kill() these
externally started dooms.
2023-02-18 22:28:15 +01:00
42a2d01f2a build tiny messaging infrastructure, refactor, hack till it sort of works
When usleep()'ing in doomsim it misses most of the quit messages. I think
it's because the controller spams RunDoom all the time and the quit
messages get discarded while doomsim is sleeping.
2023-02-18 22:08:12 +01:00
cc4f66f8de make logc an OBJECT library and enable clang-tidy for my code 2023-02-18 12:30:53 +01:00
67fba60f47 fix two "No newline at end of file" warnings 2023-02-18 12:30:53 +01:00
af41a870d6 playing with clang-tidy, needs more love 2023-02-18 12:30:53 +01:00