TDM Engine Development Page

[peter_spy]

Yup, that's exactly how it looks like.

[revelator]

tested with FBO off, its better but still has some bugs like mirror effect on some portals, with FBO on its a mess

This might take a while to fix, cause i have to find out what causes this much discrepancy between the two.

[revelator]

hmm GL_TEXTURE_COORD_ARRAY gets turned off a lot of places in the debug tools but its newer set anywhere,

i commented out all instances of it and the debug functions work the same without it, sadly by work i mean not so well :/

 

With FBO on its completely broken, and some rendering functions which where loaded internally before have been moved to shaders, so might no longer work exactly the same way anymore.

[duzenko]

hmm GL_TEXTURE_COORD_ARRAY gets turned off a lot of places in the debug tools but its newer set anywhere,

i commented out all instances of it and the debug functions work the same without it, sadly by work i mean not so well :/

 

With FBO on its completely broken, and some rendering functions which where loaded internally before have been moved to shaders, so might no longer work exactly the same way anymore.

That needs to be replaced with qglDisableVertexAttribArray( 8 ) but I don't think it's related to invalid geometry.

 

[revelator]

Indeed, it seems to be related to portals and fogshaders, as an experiment i tried to run the portal debug tool and whenever i hit a fog leaf portals would dissapear Oo so ouch.

Im not sure why it there does not seem to be any problems outside the debug render tools though.

[duzenko]

Indeed, it seems to be related to portals and fogshaders, as an experiment i tried to run the portal debug tool and whenever i hit a fog leaf portals would dissapear Oo so ouch.

Im not sure why it there does not seem to be any problems outside the debug render tools though.

How do I repeat it? I don't see this problem in e.g. Old Habits 2.

[revelator]

try enabling r_showportals together with r_showtris, it works then but why.

[revelator]

 

How do I repeat it? I don't see this problem in e.g. Old Habits 2.

 

Might not show up in the same spot in all maps, try running the gatehouse you will get the problem with r_showtris on just after entering the gate over the drawbridge.

[revelator]

Ok it works fine besides a little flickering in the portal outlines when crossing a fogleaf when r_showportals are on, tbh im a bit baffled by this.

[revelator]

Spoke to soon it seems.

 

 

With FBO on it still causes a mess :S

 

 

the tri outlines dissapear at certain angles with showportals on and FBO.

 

Without FBO and showportals the outlines are drawn as inverted mirrors when i hit these angles, sadly i cannot get it to show these on screenshots.

 

It only seems to crop up whenever there are two or more portals close together or intersecting each other.

[revelator]

Oh hell, ok FBO works also if set before entering the game or after doing a vid_restart, but only if showportals are on.

Do we need another toggle for FBO states ?.

 

Im having a nasty feeling this bug is caused by a broken state change somewhere.

[duzenko]

Sorry, I still can't see it.

I started the Gate house and passed a couple of doors and it just works for me.

[peter_spy]

When you walk to a portal and you don't see the glitches, try going through it and seeing it from the other side. Edit: using r_showportals does seem to resolve the mirror and rainbow glitches effect though.

Edited August 16, 2018 by Judith

[revelator]

Strange if it only affects some of us ?, but it does not seem to be limited to nvidia cause i use an AMD R9 390 and i get it to.

do you use an intel card duzenko ?

[peter_spy]

I think I've found the culprit. Try toggling Postprocessing on and off while looking around with showtris. When you turn postprocessing off, the mirrors and rainbows go away too.

[revelator]

Great might be a conflict in the post process code screwing up then.

The reason i was apprehensive about why showportals also seems to fix this is that it uses a totally different rendering chain (intermediate vs VBO), so it should actually cause more problems not less.

[revelator]

	if ( !w || !h // this has actually happened
		// nbohr1more add checks for render tools
		|| r_showLightCount.GetBool()
		|| r_showShadows.GetBool()
		|| r_showVertexColor.GetBool()
		|| r_showShadowCount.GetBool()
		|| r_showTris.GetBool() // revelator: need this one as well, screws up showtris as pointed out by judith.
		|| r_showTexturePolarity.GetBool()
		|| r_showTangentSpace.GetBool()
		|| r_showDepth.GetBool() ) {
		return;
	}

Ok now were cooking with gas

[revelator]

I suspect its the depthtest from the bloom code that interferes badly with the showtris code.

But we dont need bloom or other fancies when using those tools, so turning these parts off should fix a good deal of the problems.

[revelator]

Added the code above and it fixed it completely nice find judith.

[nbohr1more]

Yeah, I was trying to be conservative and only disable bloom where I absolutely knew it broke things.

[revelator]

All changes upped, it should be in the next build.

[revelator]

No problem m8, i should have expected that this was what caused it since most postprocessing adds another layer to the renderer.

[revelator]

Btw. i moved the postprocessing code into tr_backend.cpp since it was only called from there.

Unless you plan on using it outside the backend commands i could remove the typedefs for it completely and save us some space, but i wanted to hear your plans first.

[nbohr1more]

I'll let Duzenko chime in there.

 

As I recall, there was this strange stub function on the game side for activating the cvar which is why

I was having a little trouble with some of the fix mechanisms for initialization operations like reloadImages.

 

In my mind r_postprocess shouldn't be doing anything on the game side and should work like all other render cvars.

 

I guess the some of this game side design stems from the use of r_postprocess with effects like influence vision (scripts)

for when the player is cursed, etc. They wanted modders to have more SDK access to these effects

so they moved some rendering control to the game side but it really doesn't belong there in Doom 3 GPL branches.

[revelator]

Ah yeah, i was wondering about that, but the mappers might probably dissagree if we move it to a place where they do not have access in case they want it for some special effect ?.

Well looks like we need to live with this sort of hacking about then,

it's not overly intrusive now so ill leave a comment about it so that further changes to postprocessing code also warns about the posibility of breakage in other parts of the chain.

 

I also added r_showportals to the check and made a small utility function to turn off postprocessing in case we use the render tools.

/*
=============
RB_CheckTools

Check for any rendertool and mark it
because it might potentially break with post processing.
=============
*/
static bool RB_CheckTools( int width, int height ) {
	// this has actually happened
	if ( !width || height ) {
		return true;
	}
	// nbohr1more add checks for render tools
	// revelator: added some more 
	if ( r_showLightCount.GetBool() || 
		r_showShadows.GetBool()	|| 
		r_showVertexColor.GetBool() || 
		r_showShadowCount.GetBool()	|| 
		r_showTris.GetBool() ||
		r_showPortals.GetBool() || 
		r_showTexturePolarity.GetBool()	|| 
		r_showTangentSpace.GetBool() || 
		r_showDepth.GetBool() ) {
		return true;
	}
	return false;
}

which is then called in the bloom code like this

	// revelator: added small function to take care of nono's
	if ( RB_CheckTools( w, h ) ) {
		return;
	}

atm its a static function but it could be made global in case that is ever needed.