[Plagman]

I suggest you thank the HRP contributors instead, they probably had more to do with it.

[Helixhorned]

Gambini, on 12 May 2013 - 01:04 PM, said:

Changelog reads:

Quote

Classic rotatesprite: add assertion failing with DNF mod, comment another.

I wonder what that is. Can someone explain me in neanderthal words what that does?

This means that I added a guard that terminates the debug build of EDuke32 if a particular code path is taken that would otherwise lead to undefined behavior. One of these was subsequently fixed, but another remains. Classic rotatesprite doesn't handle large zoom values well because pretty much everything is calculated in fixed point. I added some instructions on how to hit the assertion failures, for example the one I fixed reads:

Quote

// Assertion would fail with DNF mod without (uint32_t) below: in mapster32,
// set dt_t 3864 (bike HUD, 700x220)
// set dt_z 16777216
// <Increase yxaspect by pressing [9]> <-- CRASH!
// (It also fails when wrecking the bike in-game by driving into a wall.)

16777216 is zoomed in 256x, but you'll notice that for the bike HUD, incorrect drawing will already start from a zoom value of about 570000 (8.7x).

So in short: The bike mod exposed some badnesses in the classic renderer, and one on those was fixed.

[DavoX]

Great helixhorned!

[TerminX]

Working on a Polymost enhancement:

[Gambini]

So awesome we´re back to the roots!

[NightFright]

Is it just me, or is EDuke32 improving even faster than usually lately?

[Micky C]

TerminX, on 13 May 2013 - 12:36 PM, said:

Working on a Polymost enhancement:

How does that work anyway?

[TerminX]

I committed it. It's still pretty experimental, but you can enable it with r_usetileshades 1. It increases memory usage for Polymost quite a bit because it requires a separate texture be generated and stored corresponding to each shade level that needs to be drawn, but it shouldn't make each .art texture take up any more memory than the average HRP texture.

[Jimmy]

Those burgundys and maroons make him look so much more menacing in the dark. Good work, I say.

[Micky C]

TerminX, on 14 May 2013 - 06:27 PM, said:

I committed it. It's still pretty experimental, but you can enable it with r_usetileshades 1. It increases memory usage for Polymost quite a bit because it requires a separate texture be generated and stored corresponding to each shade level that needs to be drawn, but it shouldn't make each .art texture take up any more memory than the average HRP texture.

Wow, I just tried this out with my WIP DNF map, and the difference is huge! Thanks!
The sky is finally the correct orange, and the rocks have much better colour variety, making them look more interesting.

[supergoofy]

R3768 - Attrition mod crashes


something is wrong with my system. I will do a restart asap and will edit my post


[edit]

it was something on my end

[LeoD]

Quote

Revision: 3759
Author: terminx
Message: Remove md4 library, since we aren't using it anywhere anymore

Oops. On Helixhorned's proposal I chose MD4 as the map identification for the (hopefully upcoming) user map maphack system.

[TerminX]

If anyone ever gets around to doing that, they can add md4 back in (or choose something better) at the time. Or, something like crc32 (which is used everywhere else) could be used instead since it's not really about security against intentional modifications, but simple identification of files. Or, maybe we can just pull in something like md5 and use that for everything everywhere... who knows.

[supergoofy]

md5 is better

[TerminX]

Sometimes speed is more important than being secure against malicious intentions; e.g we don't have much of a reason to care about someone being able to modify a GRP file to be corrupt but still show up as valid in the startup window but we do care if suddenly the GRP scan at startup takes 5 times as long.

[Fox]

Fox, on 01 May 2013 - 07:40 AM, said:

By the way, is it possible to add SoundToggle, FXVolume, MusicToggle and MusicVolume to the avaiable userdef structures?

Any light on this? I don't want to sound like a douche, but since the source code already perform a ud check on this, I think it shouldn't be too hard to implement.

[TerminX]

I would be OK with supporting reading the values, but definitely not with setting them from within CON.

[TerminX]

Another couple of comparison shots from Polymost:

[Plagman]

E1L5 is great for that stuff. I used this one tile both for highpal and artmapping testing since it has so much variance.

[TerminX]

Indeed. On that note, I'm so glad we're able to reproduce the coloring of the original game accurately (for the most part) in all renderers now.

Please test the new mode, guys... I plan on making it the default sometime soon (just like the one in Polymer is default now), assuming major problems with it don't crop up.

[Gambini]

I have been testing it a lot, along with new polymer´s shading system.








They´re both dreamlike. There are some minor problems with each one, but most of those are known problems that have been happening long before.

It´s important to remark that the relevance of these issues in comparision to the improvements is so small that it almost makes no sense to mention them. With the exception of a few polymer specific issues, which should be IMO high priority.

Visibility in polymost may need some minor tweaking:







Here you can see that in both 32bits renderers vertical offset in viewalligned sprites is mirrored:





Some gradient ramps in duke´s palette have poor shadetables transitions, this is more noticeable in Polymost now since the visibility is based on open-gl´s regular shading system, no big deal, just pointing it out (I don´t consider it worth being fixed):




Here you can see how Polymer still has problems when rendering maskwalls that are inside a double-parallaxed sector:

FORGOT TO UPLOAD THOSE SCREENS

Here you can see how both 32bit renderers draw non-square-of-2 textures wrong (old problem anywway):





Here you can see how in Polymer decal sprites still clip their background;



Not pictured problems: Grills fences and other sprites with alhpa parts become thick and blur on a relatively short distance.

[Fox]

TerminX, on 15 May 2013 - 01:42 PM, said:

I would be OK with supporting reading the values, but definitely not with setting them from within CON.

I only intend to read it, but I suppose anyone should use it at their own risk.

By the way, is ud.config.MusicDevice even used in Eduke today?

[Diaz]

Gambini, on 15 May 2013 - 07:57 PM, said:

Not pictured problems: Grills fences and other sprites with alhpa parts become thick and blur on a relatively short distance.

Yeah, to fix this I suggest implementing a way to define tiles as nomipmap, maybe in a separate file that could then be included with EDuke32 (just like tiles.cfg).

The bad thing about the other issues is that fixing them will probably break existing Polymer maps (since they have been done with those deficiencies being present). I guess, however, that fixing thousands of non-polymer usermaps and Duke itself is more important

While I'm at this, now that it seems like the priority is to get all renderers to look like classic as much as possible, may I ask that r_usenewshading value of 1 is never removed. Pretty pretty please, it would break all I've done so far

[Plagman]

I don't think there are any plans of removing any existing functionality such as this, backwards-compatibility is understood to be extremely important. Isn't anisotropic filtering supposed to mitigate things like you describe?

[Jblade]

I just played with it a little yesterday and it's really impressive stuff TerminX, I didn't really notice any performance decreases at all. It's definitly fixed one of the biggest problems I had with polymost.

[Diaz]

Plagman, on 15 May 2013 - 11:27 PM, said:

I don't think there are any plans of removing any existing functionality such as this, backwards-compatibility is understood to be extremely important. Isn't anisotropic filtering supposed to mitigate things like you describe?

Mmmmm, you're right. I don't notice grates becoming blurry with anisotropic filtering indeed. I mentioned the nomip thing off the top of my head because that's how it used to be in the past, with the Q2 and Q3 engines (might still be useful for those with videocards not fast enough to handle aniso)

[Mia Max]

Hallo!

Can we have an option for guns to prevent flash lights from fireing guns?
In huge areas that flash can ruin all the atmosphere.

This would be an great option (for me )

[Diaz]

Alright, since we can't change a sectoreffector's cstat via CON, I was wondering how hard would it be to implement a visibility check for SE 50's.

Shadows seem to decrease performance even if the SE they come from is not in view, so maybe the SE's could have cstat 64 added (so they won't cast shadows) when they are not within the player's PVS. This would help shadow performance tremendously.

[Fox]

Mia Max, on 16 May 2013 - 11:58 AM, said:

Hallo!

Can we have an option for guns to prevent flash lights from fireing guns?
In huge areas that flash can ruin all the atmosphere.

This would be an great option (for me )

Yes, you would need to add flag 256 to WEAPONx_ FLAGS.

[Danukem]

Fox, on 16 May 2013 - 12:24 PM, said:

Yes, you would need to add flag 256 to WEAPONx_ FLAGS.

He wants the option to be in the EDuke32 menu. I think it's reasonable to want a menu option, since the flashing looks bad in many vanilla maps (e.g. where fog is used).