Beta Testing 2.09

[cabalistic]

@Araneidae Did some research, and it turns out support for our 64 bit color precision is actually core since OpenGL 3.0. This is probably a driver bug - I'm afraid you'll have to remember to stick to 32 bits color precision as a workaround. There's nothing for me to detect missing support, unfortunately.

[MirceaKitsune]

I know this can be a matter of personal feeling, but at this point it feels off enough to ask: Did any cvars relating to AI difficulty change in recent versions, or perhaps their functionality altered in 2.08 or 2.09? My difficulty level is set to Nearly Deaf / Nearly Blind, yet the AI reacts almost immediately and will reach an alert level for the smallest reason. Asking here since I feel it was a bit less reactive before so I'm imagining something must have changed, either in the scripts or outdated settings perhaps.

[stgatilov]

8 hours ago, cabalistic said:

@Araneidae Did some research, and it turns out support for our 64 bit color precision is actually core since OpenGL 3.0. This is probably a driver bug - I'm afraid you'll have to remember to stick to 32 bits color precision as a workaround. There's nothing for me to detect missing support, unfortunately.

Technically, it is possible to ask OpenGL after FBO creation if FBO indeed has the desired format. But I'm not sure if it is worth the trouble, and if it would help at all.

Also, I'm not sure about the referenced news article: will this bug be fixed soon by driver update?

[cabalistic]

5 hours ago, stgatilov said:

Technically, it is possible to ask OpenGL after FBO creation if FBO indeed has the desired format. But I'm not sure if it is worth the trouble, and if it would help at all.

Also, I'm not sure about the referenced news article: will this bug be fixed soon by driver update?

Is an FBO allowed to take a different internal format than the requested one? Because we do validate them, so a failure to create the FBO as requested should at least give a warning in the console... But either way, I agree it doesn't seem worth the effort.

[Araneidae]

14 hours ago, cabalistic said:

@Araneidae Did some research, and it turns out support for our 64 bit color precision is actually core since OpenGL 3.0. This is probably a driver bug - I'm afraid you'll have to remember to stick to 32 bits color precision as a workaround. There's nothing for me to detect missing support, unfortunately.

That's annoying!

6 hours ago, stgatilov said:

Technically, it is possible to ask OpenGL after FBO creation if FBO indeed has the desired format. But I'm not sure if it is worth the trouble, and if it would help at all.

Also, I'm not sure about the referenced news article: will this bug be fixed soon by driver update?

It'd be great to pin this down to an actual identifiable bug and push this upstream ... can either of you create a tiny test case that would demo the issue?  Afraid I have absolutely zero OpenGL experience, and I understand it's potentially a lot of hassle for you with negligible benefit ... though I wonder how many other graphics cards are affected by this?

I too was unclear what the Phoronix article was actually saying!

[cabalistic]

I'm afraid a small test case isn't trivially possible, particularly since I still don't understand 100% what is going wrong for you, exactly. It isn't even as simple as saying that your card/driver doesn't support the half-float format, because when you first disabled r_tonemap, but left r_fboColorBits at 64, you were still rendering to a 64 bit framebuffer, and that seems to work fine.

The difference between the two FBOs in question is that one is using a render buffer and the other is using a texture. So I guess the problem is more related to half-float textures than render buffers, but it's hard to say for certain.

[MirceaKitsune]

I replayed Snowed Inn last night, definitely one of my favorite FM's. I remembered it has quite a few improved assets: Better footstep sounds, loot sounds, and other improved and more realistic audio overrides. Also a better looking objectives screen with slightly higher resolution graphics.

I really must ask: Unless there are licensing issues involved, is there a reason why we don't make those vanilla and replace the old ones? They're in the same style but better and more accurate! It could be a nice addition to the new 2.09 release

[MirceaKitsune]

A couple of bugs I found today. First of all I happened to replay the FM Special Delivery this time. This FM will crash and is unplayable with later TDM versions, if in the store at the start of the game you choose to buy the Speed Potion. It seems to be due to its scriptobject not being found... maybe it would be a good time to add the finished version of this potion to core?

Second one, and I noticed this since the first 2.09 beta: The red and green glows when using items appear really sharp, as if in lower color resolution. Unfortunately I forgot to turn off the sharpen filter and see if that's causing it.

Lastly I recently replayed Snowed Inn as mentioned above. Saw a little peculiarity which might be an alpha drawing issue in the renderer: There's a certain drawing on a wooden wall in a part of the map. When shining the flashlight at it, the corners of the decal plane become black rather than staying transparent. This only happens if the flashlight is on and perhaps shining at the right angle, without it transparency works just fine. Might count as a FM spoiler so just in case.

Spoiler

 

 

[Dragofer]

@MirceaKitsune the speed potion script had problems that made it have unpredictable consequences for movement in real missions, and it looks like (some of) the assets were included by mistake. If you or somebody else can find and iron out those bugs, then the scriptobject can be added to the game.

For the 2nd visual glitch, I think it'd be very helpful to check if you see the exact same thing on TDM 2.08 or earlier. This gives a major clue on where the problem lies: either with one of the recent 2.09 changes, or an old problem that's been around for a while.

[lowenz]

Of course it was in 2.08.

[PranQster]

On 1/4/2021 at 7:06 PM, MirceaKitsune said:

When shining the flashlight at it, the corners of the decal plane become black rather than staying transparent. This only happens if the flashlight is on and perhaps shining at the right angle, without it transparency works just fine.

This reminds me of a similar glitch in siegeshop for a long time. In that case it was a decal to blame. The drawing of the catapult on the nearby desk was sitting on top of a dirt decal which was on the desk surface. At the time I believe I had stacked the patches 0.125 units above the desk and each other. From a certain distance and angle, in light, the dirt decal was rendering over the paper. 

[nbohr1more]

For Windows users, please confirm whether "image_preload 0" works or whether it produces no 3D render on mission startup.

Thank you!

[AluminumHaste]

Works for me, it's been on.

[wesp5]

Another report that was already valid for the last few versions and should better be fixed in 2.09:

After downloading missions both these missions and the "nothing new..." text are displayed on top of each other!

[nbohr1more]

46 minutes ago, AluminumHaste said:

Works for me, it's been on.

Hmm...

Can you post your darkmod.cfg ?

[nbohr1more]

OK, I've narrowed it a bit more.

"image_preload 0" only breaks rendering with SSAO enable or Bloom enabled.

[D.Brown]

just downloaded it. i dont know why, but the game looks dirty now, nearly like the necro age for thief. i like it alot... especially with brightness on 1.5, gamma below 1, desaturation at ~0.5. it feels like a whole new game to me. 

thank you

edit:
these settings does not work on every map, since every other map got different lightning strength. however the new version is a blast.

Edited January 9 by D.Brown

[MirceaKitsune]

I noticed the black transparency bug on another decal again, FM Sneak & Destroy. getViewPos included this time.

Spoiler

 

[AluminumHaste]

On 1/7/2021 at 10:37 PM, nbohr1more said:

OK, I've narrowed it a bit more.

"image_preload 0" only breaks rendering with SSAO enable or Bloom enabled.

Also have both of those on.

EDIT: Only thing I have off is bindless textures due to the green tinged driver bug.

 

[Dragofer]

Started 2.09b2 on my desktop PC and found that Ambient Occlusion is broken, while it's working fine in 2.08. Both versions are freshly updated with fresh darkmod.cfg. The higher the AO setting, the more noticeable the issue becomes. Other graphics settings in the GUI menu didn't seem to have an influence.

I tweaked my r_GLcoreprofile setting and restarted the game, but that only slightly changed the manifestation of the bug:

r_GLcoreprofile 2 (default):

Spoiler

Small granulated horizontal lines where there should be AO:

 

r_GLcoreprofile 0 or 1:

Spoiler

Long granulated horizontal lines spanning across the whole width of the screen. No sign of AO:

My GPU is an AMD Radeon RX 460, my CPU an i5-6402P, Windows 10.

I've attached condumps of each glcoreprofile setting:

glcoreprofile2.txtglcoreprofile1.txtglcoreprofile0.txt

As well as darkmod.cfg's for 2.08 and 2.09b2:

Darkmod_208.cfgDarkmod_209b2.cfg

For what it's worth, my office laptop also has problems with AO, but there it also occurs in 2.08.

[nbohr1more]

9 hours ago, Dragofer said:

Started 2.09b2 on my desktop PC and found that Ambient Occlusion is broken, while it's working fine in 2.08. Both versions are freshly updated with fresh darkmod.cfg. The higher the AO setting, the more noticeable the issue becomes. Other graphics settings in the GUI menu didn't seem to have an influence.

I tweaked my r_GLcoreprofile setting and restarted the game, but that only slightly changed the manifestation of the bug:

r_GLcoreprofile 2 (default):

  Reveal hidden contents

Small granulated horizontal lines where there should be AO:

 

r_GLcoreprofile 0 or 1:

  Reveal hidden contents

Long granulated horizontal lines spanning across the whole width of the screen. No sign of AO:

My GPU is an AMD Radeon RX 460, my CPU an i5-6402P, Windows 10.

I've attached condumps of each glcoreprofile setting:

glcoreprofile2.txt 15.06 kB · 0 downloads glcoreprofile1.txt 15.02 kB · 0 downloads glcoreprofile0.txt 14.84 kB · 0 downloads

For what it's worth, my office laptop also has problems with AO, but there it also occurs in 2.08.

This is a long shot but perhaps we will see something if you reload GLSL after the mission loads.

Please invoke reloadGLSLprograms and capture a condump.

Also, please attach your darkmod.cfg file.

I am guessing that we would need to do an API trace:

https://apitrace.github.io/#download

 

[Dragofer]

@nbohr1more Thanks, I've tried reloadGLSLprograms and attached the condump (no visual difference):

reloadGLSL.txt

I've attached the darkmod.cfg's to my previous post.

I remember doing an API trace before, I'm guessing I need to follow some specific instructions?

[nbohr1more]

As I recall, to capture the API trace you install the program then invoke:

apitrace --api gl C:\darkmod\thedarkmodx64.exe --output C:\apilog.txt

Before doing this, let's try enabling:

r_debugGLSL 1

Then loading your mission and performing reloadGLSLprograms and save a condump.

[Dragofer]

Ok I've tried to go back to some earlier builds using the new tdm_installer to find roughly when it broke. Interestingly, all dev builds worked fine, and now that I'm back to 2.09b2 AO works fine even there. So I guess it was just a problem with the tdm_installer.exe

[nbohr1more]

Hmm. I wonder if shader cache from a previous build was not refreshed so the driver was trying to execute old shader data?