Beta Testing 2.08

[PranQster]

Revisiting the issues I had with sound due to issues with pulseaudio.

Currently, audio is working for the 64/32 bit Linux client due to having set the following in Darkmod.cfg:

seta s_device "HDA Intel PCH, ALC1150 Analog (CARD=PCH,DEV=0)"

This allows TDM to use alsa for output, but pulseaudio does not see this output and I have no options for muting or controlling volume for TDM through the pulseaudio volume control or plasma widgets (mute icon added to the task icon).

If I have s_device "default", I get no output to the speakers. But pulseaudio recognizes the tdm alsa plugin as an application and lists it in volume control. The tdm icon in the plasma task manager has its mute icon available, when not for the named card.

Ironically, if I run the TheDarkModx64.exe using wine with s_device "default", audio works fine as well as integration with pulseaudio. But I'd still like to get it working properly as s_device "default" with the linux binary.

I can live with it as it is, but I love having the easy option to mute the single application. That way I can mute tdm during dmap, but still have audio for darkradiant for setting up speakers etc. Currently, to mute TDM, I have to turn the master volume for my speakers down. Also, with pulseaudio, I can usually have audio for both TDM and DR. Without pulse, only the program to use audio first gets audio output. So if TDM gets launched first, only TDM gets audio. Or if a speaker is set up in DR and audio played during that process before launching TDM, then TDM gets no audio unless DR is first closed.

So far I have seen no evidence that changing anything at all in alsoft.ini, or even deleting it, has any effect whatsoever on audio with the linux binary. So if any solutions can be found by editing only the cvars in Darkmod.cfg, that would be ideal.

Edited April 20, 2020 by PranQster

[PranQster]

Elevator use seems to have issues in 2.08

I just dmap'd and loaded up siegeshop last night to check with 2.08 compatibility. Aside from a water splash particle I need to replace, I found that when two AI try to use an elevator at the same time, they come to a standoff at the call buttons. The AI just sit there doing nothing. I need to use one of the call buttons to move the elevator. Then one of the AI will finally decide to use it and everyone gets to pathing again.

To clarify, I saw this happen twice, and at each of two elevators. In both situations, the AI trying to use the elevators were on different floors from one another.

Edited April 24, 2020 by PranQster

[grayman]

Using Beta 4.

I thought this HOM problem was fixed.

View is from the eastern void in "Accountant 1".

 

[grayman]

On 4/23/2020 at 8:47 PM, PranQster said:

Elevator use seems to have issues in 2.08

I just dmap'd and loaded up siegeshop last night to check with 2.08 compatibility. Aside from a water splash particle I need to replace, I found that when two AI try to use an elevator at the same time, they come to a standoff at the call buttons. The AI just sit there doing nothing. I need to use one of the call buttons to move the elevator. Then one of the AI will finally decide to use it and everyone gets to pathing again.

To clarify, I saw this happen twice, and at each of two elevators. In both situations, the AI trying to use the elevators were on different floors from one another.

Do you have a savegame from a point just prior to the problem occurring?

If so, I'll take a look.

If not, perhaps you can create one.

Thanks.

[duzenko]

49 minutes ago, grayman said:

Using Beta 4.

I thought this HOM problem was fixed.

View is from the eastern void in "Accountant 1".

 

 

r_skipSubviews 1?

[nbohr1more]

2 hours ago, duzenko said:

r_skipSubviews 1?

I believe you had offered to force subviews off when out in the void. I guess that was never implemented.

[duzenko]

6 minutes ago, nbohr1more said:

I believe you had offered to force subviews off when out in the void. I guess that was never implemented.

Something like that, recalling vaguely

[grayman]

Suggested here.

 

[grayman]

The trunk behaves the same way, so nothing was changed since the problem was first reported.

 

[PranQster]

4 hours ago, grayman said:

Do you have a savegame from a point just prior to the problem occurring?

If so, I'll take a look.

If not, perhaps you can create one.

Thanks.

No, I didn't think of it. It would be so random that I'd need to wait for it to occur, then make a savegame while the standoff is happening. Would that work?

[grayman]

I might be able to see something, but the key thing I'm looking for is why the AI aren't queuing up properly to use the elevator. And that means seeing what's happening from the moment they realize they need the elevator to get somewhere.

[duzenko]

16 hours ago, grayman said:

The trunk behaves the same way, so nothing was changed since the problem was first reported.

 

Revision: 8683
Force subviews off when out in the void

 

[grayman]

Verified that HOM is gone from the void.

Thanks.

Pls make sure this gets merged to the 2.08 branch. 

[duzenko]

4 hours ago, grayman said:

Pls make sure this gets merged to the 2.08 branch. 

@stgatilov

[Spooks]

Hey, the alpha masking material technique that I found about circa 2016 doesn't work anymore, as far as I can tell. I'm not sure what's causing it, but the map makealpha() or the gl_dst_alpha, gl_one blending mode doesn't seem to do what it did before. Something about the change from ARB to GLSL? Dunno. Can somebody who has an earlier version of TDM (2.07, 2.06) test this one for me? Put this snippet in a material definition that's used somewhere in a map:

	{
	
		maskcolor
		map makealpha (textures/darkmod/stone/cobblestones/flagstones03_shiny_s) //here I use the specular of the texture
		alpha 1		
	}
	
	{
	blend gl_dst_alpha, gl_one
		maskalpha
		map				textures/darkmod/stone/brick/blocks_large_whitestone		
	}

In theory it should overlay a bricks texture but modulate it by the specular map of some flagstones. If it doesn't work try

{
maskcolor
map makealpha (textures/darkmod/stone/cobblestones/flagstones03_shiny_s) //here I use the specular of the texture
alpha 1		
}
	
{
blend gl_dst_alpha, gl_one
maskalpha
cubeMap      env/gen2
texgen      reflect
}

That one is a cubemap stage that I've personally used before, so I know it should be masked...

EDIT: This isn't a purely personal whim, Epifire's materials use this trick so I don't know if any of them with this thing have made it into the mod; if they have they'd be broken in a way that's very hard to spot but is important for them working correctly.

Edited April 26, 2020 by Spooks

[PranQster]

On 4/25/2020 at 12:38 PM, grayman said:

I might be able to see something, but the key thing I'm looking for is why the AI aren't queuing up properly to use the elevator. And that means seeing what's happening from the moment they realize they need the elevator to get somewhere.

It's also quite possible that this behavior has always been there, but the RIT networks make it so rare for this to occur that I simply never saw it happen before. It just happened to be with my first run with 2.08 that I saw it happen twice.

I just saved the map in DR 2.80 and dmap'd in 2.08 and so far only see issues with a watersplash particle. No conflicts yet witnessed at the elevators. If I see two AI heading for the same elevator I will make a savegame. 

[nbohr1more]

2 hours ago, PranQster said:

It's also quite possible that this behavior has always been there, but the RIT networks make it so rare for this to occur that I simply never saw it happen before. It just happened to be with my first run with 2.08 that I saw it happen twice.

I just saved the map in DR 2.80 and dmap'd in 2.08 and so far only see issues with a watersplash particle. No conflicts yet witnessed at the elevators. If I see two AI heading for the same elevator I will make a savegame. 

I doubt it's related but one thing you can try is setting "tdm_ai_search_type" to 1 (2.03 behavior) and see if you can reproduce it.

That may give you a shortcut to bisect.

[PranQster]

4 hours ago, nbohr1more said:

I doubt it's related but one thing you can try is setting "tdm_ai_search_type" to 1 (2.03 behavior) and see if you can reproduce it.

That may give you a shortcut to bisect.

Thanks. I'll try it.

I did try making a savegame, shortly after which (about 1 minute) there was a standoff between one AI exiting and another trying to enter the elevator. I loaded that savegame and moved near the elevator and saved again. But this time the AI behaved as they should.

Edited April 27, 2020 by PranQster

[nbohr1more]

13 hours ago, Spooks said:

Hey, the alpha masking material technique that I found about circa 2016 doesn't work anymore, as far as I can tell. I'm not sure what's causing it, but the map makealpha() or the gl_dst_alpha, gl_one blending mode doesn't seem to do what it did before. Something about the change from ARB to GLSL? Dunno. Can somebody who has an earlier version of TDM (2.07, 2.06) test this one for me? Put this snippet in a material definition that's used somewhere in a map:

	{
	
		maskcolor
		map makealpha (textures/darkmod/stone/cobblestones/flagstones03_shiny_s) //here I use the specular of the texture
		alpha 1		
	}
	
	{
	blend gl_dst_alpha, gl_one
		maskalpha
		map				textures/darkmod/stone/brick/blocks_large_whitestone		
	}

In theory it should overlay a bricks texture but modulate it by the specular map of some flagstones. If it doesn't work try

{
maskcolor
map makealpha (textures/darkmod/stone/cobblestones/flagstones03_shiny_s) //here I use the specular of the texture
alpha 1		
}
	
{
blend gl_dst_alpha, gl_one
maskalpha
cubeMap      env/gen2
texgen      reflect
}

That one is a cubemap stage that I've personally used before, so I know it should be masked...

EDIT: This isn't a purely personal whim, Epifire's materials use this trick so I don't know if any of them with this thing have made it into the mod; if they have they'd be broken in a way that's very hard to spot but is important for them working correctly.

Does the same happen if you set r_testSpecularFix 0 ?

[Spooks]

Hello, yes, specularFix 0 seems to do nothing with regard to the problem. Epifire's cubemap testing map could be a useful testbed (cabalistic posted about parallax cubemaps in his thread), provided one inverts the "Marble_Wall_001_mask.tga" texture so the mask is more clearly visible.

[Spooks]

Update: I'm now fairly sure that it's another .dds vs .tga issue. TGAs actually work with the makeAlpha keyword, but the thing is I'm 90% certain I've made it work with .dds textures (spec maps) before. I don't know if something's changed or I'm crazy, but at least I've narrowed the issue down to file formats.

[nbohr1more]

Image programs like "makeAlpha" don't work with DDS. Not sure about maskalpha though.

Edit: Confirmed in source. maskAlpha should work with any image type but makeAlpha has to read

RGB directly from the texture and the engine lack this feature for DDS.

 

[grayman]

Problems with newspaper readable gui:

Getting these warnings:

 

I tried tracking down the root cause, but got lost in the weeds of gui instruction/definition inclusion. Everything looked okay to me.

 

[grayman]

Getting a warning that the number of values provided for Ambient Occlusion doesn't match the number of text descriptions provided in the gui:

        choiceDef AmbientOcclusion
        {
            rect        SETTINGS_X_OFFSET, 171, 80, MM_LINE_H
            choices        "#str_07221" // Off;Low;Medium;High(;Very High)
            values        "0;1;2;3"
            cvar        "r_ssao"
            updateGroup    render
            CHOICE_DEF
            onAction { resetTime "AnimGammaDown" 0; }
        }

In all.lang we see "#str_07221"    "Off;Low;Medium;High;Very High".
        
The parser finds 5 values in the "choices" keyword, but only finds 4 values in the "values" keyword (thus the warning).

The code that processes r_ssao doesn't include anything to handle the value "4", so a "4" will default to "default", which is the same as the value "3" (High).

Could this please be corrected to make it explicitly 4 or 5 choices, whichever the intent is? Less spam = goodness.

Thanks

[Springheel]

8 hours ago, grayman said:

Problems with newspaper readable gui:

Getting these warnings:

I tried tracking down the root cause, but got lost in the weeds of gui instruction/definition inclusion. Everything looked okay to me.

 

If that's the newspaper gui I made for New Job, it isn't set up to use page corners as it's only meant to be a single page.