Rework material system material and surface generation

[VReaperV]

Some more changes to allow entities to be added to the material system. This entirely removes the GenerateWorldMaterials() function, instead materials are generated for each MaterialSurface. This also removes a bunch of things from drawSurf_t, tess, Render_*() etc. that were only used to pass information through to the material system.

Also fixed a memory leak in material system portal surface processing.

[slipher]

Ran tests and got this: ^3Warn: Unhandled exception (15ShaderException): Shader not found: deformVertexes_pulse_2 0. Note that I am NOT using the dummygame.

[VReaperV]

Is it possible it's a glsl cache issue? I've just loaded pulse with engine built from this pr and it works fine.

[slipher]

I can reproduce that with just ./daemon -set r_materialSystem 1 +devmap pulse.

[slipher]

Is it possible it's a glsl cache issue? I've just loaded pulse with engine built from this pr and it works fine.

No, I still reproduce it with a pristine homepath.

[VReaperV]

I did the same thing with homepath, still can't reproduce. I've also added some logging locally, and deformVertexes_pulse_2 0 is loaded after materials are generated.

[illwieckz]

I can reproduce that with just ./daemon -set r_materialSystem 1 +devmap pulse.

I get the exact same (AMD RDNA3, Mesa radeonsi).

[VReaperV]

I can reproduce that with just ./daemon -set r_materialSystem 1 +devmap pulse.

I get the exact same (AMD RDNA3, Mesa radeonsi).

So it works on master, but not on this branch? The only thing I could think of is that there's some end-of-string character or something that is messing it up (between tr.loadingMap and tr.world->baseName). Actually, even that shouldn't be able to cause it, because it would just load a new deform shader then.

[illwieckz]

So it works on master, but not on this branch?

Yes, on my end it works on master but not on this branch.

[VReaperV]

I was able to reproduce this on vega instead. The issue itself is present on master, but apparently it gets hidden by a race condition: the gfx/buildables/human_base/spawning also uses deform shaders, and it can be loaded before the map shaders since it comes from a QVM call, thereby hiding a bug in deform shader lookup.

[VReaperV]

Added a commit to fix this. This should also fix the similar issue in #1600.

[slipher]

My tests now complete without erroring out, but they reveal some regressions when material system is enabled. For example:

[VReaperV]

That's chasm and watah?

[VReaperV]

Can't reproduce the watah one running the regular gamelogic, I assume the above screenshots are taken with dummy gamelogic instead.

[slipher]

No. I don't really use dummy gamelogic anymore since the GLSL building changes broke it.

[VReaperV]

Any non-default graphics settings other than r_materialSystem 1?

[illwieckz]

This is what I get on Watah on viewpos -2882 3079 1285 -76 2, with material system:

Without:

I get the same on mastet anyway, so that's not this branch's fault.

[illwieckz]

I experience no regression between master and this anymore.

[illwieckz]

LGTM.