The assertion that triggered was "assert(iconvDesc == (SDL_iconv_t)-1);"
in Sys_InitInput() - because when loading a mod the window is recreated,
calling Com_ReloadEngine_f() -> idCommonLocal::InitGame()
-> idCommonLocal::InitRenderSystem() -> idCommonLocal::InitOpenGL()
-> R_InitOpenGL() -> Sys_InitInput()
Before that Com_ReloadEngine_f() calls commonLocal.ShutdownGame( true );
which calls the equivalent Shutdown() functions - except so far no one
called Sys_ShutdownInput() (which closes iconvDesc and resets it to -1).
Fixed that by making idRenderSystemLocal::ShutdownOpenGL() call it.
For some reason Wayland thought it would be clever to be the only
windowing system that (non-optionally) uses the alpha chan of the
window's default OpenGL framebuffer for window transparency.
This always caused glitches with dhewm3, as Doom3 uses that alpha-chan
for blending tricks (with GL_DST_ALPHA) - especially visible in the main
menu or when the flashlight is on.
So far the workaround has been r_waylandcompat which requests an OpenGL
context/visual without alpha chan (0 alpha bits), but that also causes
glitches.
There's an EGL extension that's supposed to fix this issue
(EGL_EXT_present_opaque), and newer SDL2 versions use it (when using
the wayland backend) - but unfortunately the Mesa implementation is
broken (seems to provide a visual without alpha channel even if one was
requested), see https://gitlab.freedesktop.org/mesa/mesa/-/issues/5886
and https://github.com/libsdl-org/SDL/pull/4306#issuecomment-1014770600
for the corresponding SDL2 discussion
To work around this issue, dhewm3 now disables the use of that EGL
extension and (optionally) makes sure the alpha channel is opaque at
the end of the frame.
This behavior is controlled with the r_fillWindowAlphaChan CVar:
If it's 1, this always is done (regardless if wayland is used or not),
if it's 0 it's not done (even on wayland),
if it's -1 (the default) it's only done if the SDL "video driver" is
wayland (this could be easily enhanced later in case other windowing
systems have the same issue)
r_waylandcompat has been removed (it never worked properly anyway),
so now the window always has an alpha chan
There were lots of places in the code that called Sys_GrabInput(),
some of them each frame.
Most of this is unified in events.cpp now, in handleMouseGrab() which
is called once per frame by Sys_GenerateEvents() - this makes reasoning
about when the mouse is grabbed and when not a lot easier.
Sys_GrabInput(false) still is called in a few places, before operations
that tend to take long (like loading a map or vid_restart), but
(hopefully) not regularly anymore.
The other big change is that the game now uses SDLs absolute mouse mode
for fullscreen menus (except the PDA which is an ugly hack), so the
ingame cursor is at the same position as the system cursor, which
especially helps when debugging with `in_nograb 1` and should also help
if someone wants to integrate an additional GUI toolkit like Dear ImGui.
glStencilOpSeparateATI() should behave exactly the same as
glStencilOpSeparate() so supporting it is easy enough and might help
some people with hardware or drivers that don't support OpenGL 2.0,
like the Mac OSX versions for PPC.
- Fix build with SDL <=2.0.3
SDL_GetGlobalMouseState was introduced in 2.0.4
(which doesn't support OSX 10.5 or older)
- Don't include execinfo.h on Mac OS X 10.4
This file isn't included in the 10.4 SDK
- Use custom typedef for PFNGLSTENCILOPSEPARATEPROC on OSX 10.4/10.5
because the system OpenGL headers for those versions don't have it
The problem was that negative values (from dhewm3tmpres.xyz) were passed
to POW, and POW doesn't have to support negative bases, according to
ARB_fragment_program.txt, and Intels Linux drive apparently doesn't,
see also https://gitlab.freedesktop.org/mesa/mesa/-/issues/5131
Using MUL_SAT instead of MUL to clamp the value that gets passed to POW
afterwards to [0, 1] fixes the problem without any disadvantages.
so far the code assumed that "result.color" is always used directly,
but the ARB shaders allow creating an alias with the aforementioned
syntax. So turn that into an variable-alias for dhewm3tmpres.
this *shouldn't* matter, but due to some Mesa bug is does:
If the shaders have been loaded already (with R_LoadARBProgram()),
then loading them again (like from the `reloadARBprograms` console cmd
or as it happens if the `r_gammaInShader` has been modified) will
cause glitches with the open source radeonsi driver (maybe also with
others? at least the open source intel driver seems unaffected).
As r_gammaInShader was marked as modified at startup (before the shaders
were even loaded) they were loaded twice: First as expected when OpenGL
is initialized, then again in R_CheckCvars() which is executed each
frame. Marking as at not modified in R_InitOpenGL() prevents this and
thus works around the bug.
However this means that changing r_gammaInShader at runtime will still
trigger this bug (while with non-broken drivers it switches seamlessly
between gamma in shader and gamma in hardware without a vid_restart).
only override cmd->viewDef in RB_DrawView() if we're drawing the
primary view (which for several calculations before actual drawing
was set to the saved/locked render view)
Note that r_lockSurfaces is more useful with r_useScissor 0 (otherwise
there's black bars over the screen when moving) and r_shadows 0 (otherwise
areas that weren't visible when locking are black because the lights
there are skipped)
remaining bug: gui surfaces move around the screen when looking around
The problem was that the editors called ChoosePixelFormat() instead of
wglChoosePixelFormatARB() - and the normal ChoosePixelFormat() has no
attribute for MSAA, so if MSAA is enabled (by SDL2 which calls the wgl
variant), ChoosePixelFormat() will return an incomaptible format and
the editors don't get a working OpenGL context.
So I wrote a wrapper around ChoosePixelFormat() that calls the wgl variant
if available, and all the necessary plumbing around that.
While at it, removed the unused qwgl*PixelFormat function pointers and
supressed the "inconsistent dll linkage" warnings for the gl stubs
the commented out code clamped the loaded filesize of affected .dds
images (crash observed with dds/guis/assets/splash/pdtempa.dds) to
200KB - but then later tries to load it and skip the first mipmap,
resuling in reading invalid memory (> 200KB into the file).
No idea what this was supposed to achieve, but it's disabled now
and the crash (at startup) is gone.
fixes#374
libjpeg is a pain in the ass, especially due to Ubuntu shipping
libjpeg-turbo in jpeg8 mode as their default libjpeg, while every other
distro I checked (including debian!) ships libjpeg-turbo in jpeg6.2 mode
Thankfully stb_image.h exists - just a single header and it even
has a (much!) friendlier API.
It's not like Doom3 (or any of the mods I checked) actually use JPEGs,
but I'm sure if I'd drop support completely, someone would complain
(perhaps rightfully so).
.. that doesn't switch the display resolution, but creates a borderless
fullscreen window at current desktop resolution.
SDL2-only (SDL_WINDOW_FULLSCREEN_DESKTOP).
Doing this with an additional CVar instead of r_fullscreen 2 or similar
has the advantage that it works properly with Alt-Enter
SDL has a bug (at least on Windows) where SDL_CreateWindow() with
SDL_WINDOW_FULLSCREEN doesn't use the configured resolution (if it's
higher than the current desktop resolution).
Try to work around that - based on Yamagi Quake II code.
Also, if GLimp_Init() fails, the "safe mode" fallback is now in
windowed mode instead of fullscreen mode.
The code used #ifdef __ppc__, probably a leftover from old Mac support,
which apparently doesn't work on all PPC systems (maybe not 64bit PPC).
Now using the LittleFloat() function instead which exists for
exactly this purpose and already handles endianess properly.
I worked around the issue for the particle editor, but now it turned out
it can also somehow happen when switching from the Radiant to the Engine
(with F2), so I implemented the "proper" fix of restoring
glConfig.vidWidth/Height, that are overwritten in
RenderSystemLocal::BeginFrame(), in RenderSystemLocal::EndFrame().
turns out tri->numIndexes can be quite big so _alloca16() would be
called with >1MB - and the Win32 stack only is 1MB, so that overflows.
As a workaround, use Mem_Alloc16() if we need >600KB
was reported in #265
can be enabled/disabled with the r_useStencilOpSeparate for comparisons
(like Z-Fail, this doesn't really seem to make a difference on my main
machine, neither on my RPi4)
Partly based on Pat Raynor's Code:
2933cb5545/neo/renderer/draw_stencilshadow.cpp
Bringing back support for the (MFC-based Windows-only) editing tools.
Only tested 32bit (not sure if the code is 64bit-clean), needs a
"proper" (non-Express) version of Visual Studio *including* MFC support,
tested the free "Community" Editions of VS 2013 and 2017.
(For VS2013 I needed to install the "Visual C++ MFC MBCS Library for
Visual Studio 2013" because it either didn't include MFC support at all
or only the wrong version, don't remember, you'll need that one anyway)
In short, it uses a idGLDrawableMaterial widget that calls
renderSystem->BeginFrame(w, h); (with w and h being small for the texture
preview window) and BeginFrame() sets glConfig.vidWidth and vidHeight to
w/h and that never got reset to the original value (window width/height).
This breaks everything because for some reason
renderSystem->GetSCreenWidth()/Height() return glConfig.vidWidth/Height
so it will just continue to render everything at that resolution (in a
small rectangle on the lower left corner of the window).
This bug has already existed in Doom3 1.3.1 (but was less noticable because
apparently when switching away from Doom3 and back to the window it reset
vidWidth/Height to the window resolution)
I only implemented a workaround (restore glConfig.vid* after rendering the
texture preview), it's possible that the same issue exists in other
(probably editor-) code - but a "proper fix" might also break code (and I'm
not super-familiar with the editor code or even just using them)
When selecting a texture in Inspectors -> Media -> Textures that doesn't
really exist (.mtr defines it, but referenced image files are missing),
the game/editor used to crash.
The problem was that somehow people thought the best way to communicate
that a file wasn't found was by setting the timestamp to 0xFFFFFFFF
and then checking for that - sometimes (incorrectly) by comparing it to -1.
That worked for 32bit ID_TIME_T (typedefed to time_t), but not with 64bit
time_t, which now seems to be the default for Win32 in VS.
So I replaced a few -1 and 0xFF... with FILE_NOT_FOUND_TIMESTAMP and now it
works.
FILE_NOT_FOUND_TIMESTAMP is still defined as 0xFFFFFFFF, maybe I should
change that, unsure if that'd break anything though..
When changing it one should keep in mind that time_t might still be 32bit
on some platforms (Linux x86?) so that should still work.. (-1 could work)
...by including `vecLib/vecLib.h` only on `__APPLE__`, and not just any `__ppc__`
From https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=235668:
> On powerpc* platforms vecLib/vecLib.h is included by default.
> This header is only available on Mac OS X, so remove it. It
> doesn't seem to be actually needed to compile this port.
Editor also seems to start, didn't test much further.
Only tested 32bit Windows, I fear the editor code isn't 64bit clean..
I hope I haven't broken anything elsewhere..
So stuff doesn't look so distorted in widescreen resolutions.
Implies that there are black bars on the left/right then..
Can be disabled with "r_scaleMenusTo43 0"
Does *not* affect the HUD (incl. crosshair) - scaling it automagically
would be very hard (or impossible), because it doesn't only render
the crosshair, healthpoints etc, but also fullscreen effects like the
screen turning red when the player is hit - and fullscreen effects
would look very shitty if they didn't cover the whole screen but had
"empty" bars on left/right.
(Mostly) fixes#188 and #189
Mods that have their own video settings menu can tell dhewm3 to replace
the "choices" and "values" entries in their choiceDef with the
resolutions supported by dhewm3 (and corresponding modes).
So if we add new video modes to dhewm3, they'll automatically appear in
the menu without changing the .gui
To enable this, the mod authors only need to add a "injectResolutions 1"
entry to their resolution choiceDef. By default, the first entry will
be "r_custom*" for r_mode -1, which means "custom resolution, use
r_customWidth and r_customHeight".
If that entry shoud be disabled for the mod, just add another entry:
"injectCustomResolutionMode 0"
This is an ugly hack that allows both exporting additional functions
(incl. methods via static function + void* userArg) to Game DLLs
and setting callback functions from the Game DLL that the Engine will
call, without breaking the Game API (again after this change).
This is mostly meant for replacing ugly hacks with SourceHook and
similar and mods (yes, this is still an ugly hack, but less ugly).
See the huge comment in Common.h for more information.
Right now the only thing implemented is a Callback for when images
are reloaded (via reloadImages or vid_restart) - Ruiner needs that.
Also increased GAME_API_VERSION to 9, because this breaks the A[PB]I
(hopefully after the next release it won't be broken in the foreseeable
future)
On FreeBSD, the game used to crash when loading the last level of RoE
(d3xp), while loading models/david/hell_h7.ma.
The problem could be reproduced on Linux whith #define USE_LIBC_MALLOC 1
and clang's AddressSanitizer.
Turns out that this file specifies a vertex transform for a non-existent
vertex (index 31, while we only have 0-30) and thus the bounds of
pMesh->vertexes[] are violated.
I added a check to ensure the index is within the bounds and a Warning
if it isn't.
It should work now. If however it turns out that more files have this
problem, maybe .ma is parsed incorrectly and we need a differently fix.
(Should) fix#138
* r_mode defaults to 5 (1024x768), I think that's more sane than 640x480
* r_fullscreen defaults to 0 (=> windowed mode) because fullscreen in
the wrong resolution sucks.. let people do their initial configuration
in windowed mode
* r_swapInterval defaults to 1 (=> VSync active by default) because that
makes the game feel more smooth and most PCs should be able to 60fps
in this 11years old game anyway
* s_useEAXReverb defaults to 1 (=> use EAX/EFX effects by default),
because OpenAL-soft supports them on all platforms/hardware and if
for some reason the used OpenAL implementation doesn't support it,
it's deactivated automatically anyway.
All these things can be configured in the Options Menu.
the resolutions are really hardcoded in an ugly combination of the
values r_mode supports, a string in strings/*.lang ("#str_04222")
describing the resolutions r_mode supports
("640x480;800x600;1024x768;1152x864;1280x1024;1600x1200")
and a string in mainmenu.gui with the corresponding r_mode values
("3;4;5;6;7;8").. as neither the strings nor mainmenu.gui are GPL'ed
I can't really redistribute a changed version of them.
So I added lots of resolutions to r_vidModes and wrote two functions
that generate the resolutions list string and r_mode value
string for the GUI.
Then I added a hack in the code that detects when the "window" for the
system options ("choiceDef OS2Primary") is created and overwrites the
hardcoded strings with custom ones from my new functions.
This is tested with both the main game and the official d3xp
(Resurrection of Evil) Addon.
No idea if it works with other mods, depends on whether they just copied
that part of the menu or wrote their own.
Because Debian Squeeze's libjpeg6 didn't have jpeg_mem_src(), we added
jpeg_memory_src() to provide the functionality.
This shouldn't be needed anymore and without it we can drop libjpeg code
from our repo.
Fixes#110
GCC had shitloads of superfluous warnings wherever List.h and Str.h were
included.. get rid of them by using #pragma GCC diagnostic at some places
in List.h and Str.h.
Also add some casts, initialize some variables for other warnings
* the OpenAL function definitions mustn't include __declspec(dllimport)
=> fixed by pretending to compile OpenAL statically
* glimp.cpp shouldn't be used in dedicated-only mode (as it was already
the case on Linux and OSX)
=> No special handling for ID_DEDICATED needed in glimp.cpp, as it's not
used anyway
* add APIENTRY to every gl function in stub_gl.cpp for compatibility
with windows headers and MSVC
* remove GL/gl.h #include from win_local.h as it's not needed
* in qgl.h, when building dedicated server for windows, redefine WINGDIAPI
to nothing for SDL_opengl.h #include to get rid off __declspec(dllimport)
by using #pragma push_macro and pop_macro, because our stub is no dll.
Fixes https://github.com/dhewm/dhewm3/issues/39
On windows un-grabbing (or possibly re-grabbing) the mouse seems to
mess up the view angly (view is centered afterwards), see
https://github.com/dhewm/dhewm3/issues/18
That's fixed by not releasing the mouse (there is no reason to do that
anyway).
In 09ea2ca8 we fixed a memset(), but it turns out wiping memory at that
position is wrong in the first place:
If an ase mesh has a GEOMOBJECT/NODE_TM transform matrix before the
GEOMOBJECT/MESH entry, the prior parsed TM will be wiped.
Since a GEOMOBJECT aseObject_t is already calling memset() upon
initialisation, we can just drop the spurios call.
Fixes#13.
This happens before the SDL GL window is set up, which just
results in a warning. Another call will properly set the
gamma table after the window is initialized.
Make gammaTable a stack variable. This is sufficient, since SDL
already restores the gamma values for us on exit.
Found another copy in renderer/Cinematic.cpp. Move the
newer implementation from renderer/Image_files.cpp in its
own file and use the libjpeg v8c license blob (which is where
this code comes from).
Adapt README too with the v8c license.
This function only exists starting with libjpeg8.
Check for existence and provide an implementation if necessary.
This allows libjpeg6 to be used.
Code borrowed from libjpeg8, adapt copyright in README
accordingly.
The class is only used for debugging and statistical purposes.
The precision is now reduced to milliseconds, but that's only
relevant for fine grained debug timings - where the old code
was inaccurate at anyway.
Don't include the lazy precompiled.h everywhere, only what's
required for the compilation unit.
platform.h needs to be included instead to provide all essential
defines and types.
All includes use the relative path to the neo or the game
specific root.
Move all idlib related includes from idlib/Lib.h to precompiled.h.
precompiled.h still exists for the MFC stuff in tools/.
Add some missing header guards.
