Quality of this commit is questionable. I have no idea whether what I did
here makes sense or not.
But it compiles and runs (provided with hlsdk-xash3d amd64 build, which
is another story).
This omits necessary waf changes that are necessary to make a valid
64-bit build. Apparently it's not enough to just pass `-8` on Windows,
you also need to hack wscript to add `x64` target to MSVC. I'll do that
later when I figure out how.
This change is a precursor for RTX Vulkan effort --
VK_KHR_ray_tracing_pipeline and friends are only available on 64-bit
nvidia drivers (no idea about AMD, pls send GPUs onegai).
They were negative by default but they should've been positive:
> a value ranging from -32768 to 32767. Triggers, however, range from 0 to 32767 (they never return a negative value).
https://wiki.libsdl.org/SDL_GameControllerGetAxis
This fixes LTRIGGER and RTRIGGER (L2/Zl and R2/Zr on most gamepads)
Previously R_CollectRendererNames left gaps in the `ref` subarrays.
This caused `R_Init` to fail in some cases because it iterates between `0` and `ref.numRenderers`.
The `.editorconfig` file in this repo is configured to trim all trailing
whitespace regardless of whether the line is modified.
Trims all trailing whitespace in the repository to make the codebase easier
to work with in editors that respect `.editorconfig`.
`git blame` becomes less useful on these lines but it already isn't very useful.
Commands:
```
find . -type f -name '*.h' -exec sed --in-place 's/[[:space:]]\+$//' {} \+
find . -type f -name '*.c' -exec sed --in-place 's/[[:space:]]\+$//' {} \+
```
1. `libref_gles2_gl4es.so` -> `libref_gl4es.so` because this is what's
passed to `R_LoadRenderer`.
2. Add `gl4es` to `DEFAULT_RENDERERS` so that `R_CollectRendererNames`
finds it.
1. Do not disable SDL_JOYSTICK events. Disabling these events causes
game controller events to be disabled as well. Instead, filter these events out.
2. Fix button mapping (it was off by one).
This fixes fades that occur at the beginning of a map being a tad screwy as a result of the ref state not being updated with fresh values, which only happens when a frame is being rendered (loading happens to prevent that).
This fixes a minor bug where sounds played via pfnPlaySoundByNameAtLocation (mainly from viewmodel animations e.g. MP5) would be left where it was emitted in the world, rather than following the local player as it does in Goldsrc.
