Fix Daemon on XWayland

[necessarily-equal]

Fixes #658

[necessarily-equal]

Alternatively there is this patch

diff --git a/src/engine/sys/sdl_glimp.cpp b/src/engine/sys/sdl_glimp.cpp
index 1500e7cd..f14662d0 100644
--- a/src/engine/sys/sdl_glimp.cpp
+++ b/src/engine/sys/sdl_glimp.cpp
@@ -522,8 +522,14 @@ static bool GLimp_DetectAvailableModes()
 
        if ( SDL_GetWindowDisplayMode( window, &windowMode ) < 0 )
        {
+               // we can't use SDL_GetWindowWMInfo yet, so we can't use
+               // that and use getenv instead.
+               if (Q_stricmp(getenv("XDG_SESSION_TYPE"), "wayland") == 0)
+               {
+                       return true; // don't fail as this is normal on XWayland.
+               }
                logger.Warn("Couldn't get window display mode: %s", SDL_GetError() );
-               return true;
+               return false;
        }
 
        for ( i = 0; i < SDL_GetNumDisplayModes( display ); i++ )

[illwieckz]

By doing that you fix one issue but bring back others (including segmentation faults).

One issue is described in the code itself here:

When calling GLimp_CreateContext() some drivers may provide a valid
context that is unusable while GL_CheckErrors() catches nothing.
For example 3840×2160 is too large for the Radeon 9700 and
the Mesa r300 driver may print this error when the requested
resolution is higher than what is supported by hardware:

r300: Implementation error: Render targets are too big in r300_set_framebuffer_state, refusing to bind framebuffer state!

The engine would later make a segfault when calling GL_SetDefaultState
from tr_init.cpp if we don't do anything.
Hopefully it looks like SDL_GetWindowDisplayMode can raise this error:

Couldn't find display mode match

so we catch this SDL error when calling GLimp_DetectAvailableModes()
and force a safe lower resolution before we start to draw to prevent
an engine segfault when calling GL_SetDefaultState().

It looks like what you get is a variant of the issue described in PR #644 (now replaced by PR #657).

[illwieckz]

If we're gonna make this test return true, we better revert the whole function to a void one.

The purpose of this function being a bool one and this test returning false was to catch this precise “Couldn't find display mode match” error and to prevent a segfault (so it was 100% done on purpose to return false there and abort).

@necessarily-equal What happens if you run latest master on Wayland, including the just merged:

• sdl_glimp: also recreate window on resolution change #657

Does it return to a safe resolution?

Note that this is not a fix for the issue you experience, but it would help to understand more what's going on.

[illwieckz]

@necessarily-equal can you add this simple comment above the return true; line

/* FIXME: returning true means the engine will crash if the window size is larger than what the GPU can do,
but we need to not fail to open window that are not using native screen resolutions even if the GPU can handle it. */

I assume it's better to give everyone the ability to open a window with an arbitrary size than preventing a crash that is only likely to happens on some obsolete hardware if the user tries to use an higher resolution than his GPU can do (which is expected to be already larger than his screen, so not something he wants).

We can do better things after 0.53, 0.52 didn't catch the said crash so not catching this crash in 0.53 isn't a regression, while preventing the creation of arbitrary window sizes is a regression.

[necessarily-equal]

Sounds good. I'll be afk tomorrow, you can push this to the PR and merge it if you want to proceed with the release Le 27 juillet 2022 19:00:47 GMT+02:00, Thomas Debesse ***@***.***> a écrit : ***@***.*** can you add this simple comment above the `return true;` line

…

``` /* FIXME: returning true means the engine will crash if the window size is larger than what the GPU can do, but we need to not fail to open window that are not using native screen resolutions even if the GPU can handle it. */ ``` I assume it's better to give everyone the ability to open a window with an arbitrary size than preventing a crash that is only likely to happens on some obsolete hardware if the user tries to use an higher resolution than his GPU can do (which is expected to be already larger than his screen, so not something he wants). We can do better things after 0.53, 0.52 didn't catch the said crash so not catching this crash in 0.53 isn't a regression, while preventing the creation of arbitrary window sizes is a regression. -- Reply to this email directly or view it on GitHub: #659 (comment) You are receiving this because you were mentioned. Message ID: ***@***.***>

[illwieckz]

merged as ba35389