[MetHy]

Hey, me again

I've already mentioned this glitch in another thread and I found a way to 'fix" the problem but the 'fix' I found doesn't work anymore.

Here is the thing, some textures get misaligned when switching to 32bit mode.
Check the metal texture against the pavement at the bottom, here is how it looks in 8bit (how I built it) :
http://i101.photobuc...zps49113ec7.png

and here is how it looks in 32bit:
http://i101.photobuc...zpsd51a4eaf.png

That's the 3rd time this happens to me, I used to 'fix' this issue simply by pasting a non glitchy texture (in this example, the same texture on the right doesn't glitch) onto the glitchy ones; however that doesn't even work anymore, or at least not in this case.

Any idea ? That trimming texture is actually around almost all of this map and now it's glitchy most of the times in 32bit so it's pretty bad in this case for me...
Joined is an example map with the glitch occuring. All I know is that it's "32-bit" specific, and not just Eduke32, it also appears glitchy in Megaton. Meanwhile I'll try to find another way to get around the issue...

Edit: forgot to mention but obviously, if you align those textures so that they look right in 32bit, it will be the contrary and they'll look off in 8bit

Attached File(s)

•  ALIGNGLITCH.rar (575bytes)
  Number of downloads: 368

[Mark]

Helixhorned, on 24 May 2014 - 10:08 AM, said:

Actually, there's a workaround for Shift+KP5+KP2/4/6/8 since r1788: just use Alt insead.

That works. Thanks for the info.

[MetHy]

MetHy, on 24 May 2014 - 01:20 PM, said:

Hey, me again

In r4482 in mapster, sometimes when I left-click to select a sprite in the textures list (the list of all textures), it actually gives me the sprite next to the one I chose rather than the one I want. Sometimes it happens, sometimes it doesn't, but it seems to happen whenever I just used the mousewheel to go through the list.

Edit : also, now the helicopter ambientsound (ambient sound#279) has much lower volume (or radius?) than before

[Daedolon]

MetHy, on 24 May 2014 - 01:20 PM, said:

Here is the thing, some textures get misaligned when switching to 32bit mode.

32 bit modes always draw textures accurately and 8 bit is to blame (or rather the textures) for not being power of two.

MetHy, on 25 May 2014 - 12:04 PM, said:

In r4482 in mapster, sometimes when I left-click to select a sprite in the textures list (the list of all textures), it actually gives me the sprite next to the one I chose rather than the one I want.

Another SDL 2.0 problem (same category as rAlt etc), been pissing me off ever since the SDL switch.

[MetHy]

Daedolon, on 25 May 2014 - 03:23 PM, said:

32 bit modes always draw textures accurately and 8 bit is to blame (or rather the textures) for not being power of two.

Then how comes the same texture used in the same manner sometimes will look off and sometimes not ? (that's what happens in my example) I also wonder why the fix that I had found no longer works...The only solution I've thought of is replacing that texture by one alike that doesn't glitch in 32bit but that one looks a bit worse so I'm tempted to say "fuck you 32bit" and leave the glitchy stuff in the map (I also thought of using a hundred sprites all over the map that would go over the glitchy textures, but due to the alignment for the texture i used I can't do that).
Btw "8bit" being to blame when everything looks right in 8bit doesn't make much sense to me. If it looks right in DOS Duke but doesn't in 32bit, whatever the issue comes from, I'll take that as a problem with the 32bit renderer.

Edit : it seems that the reason why I can't fix the glitchy textures simply by pasting the same non glitchy texture onto it, is because in this case the glichty textures are at the bottom of a texture swap bit (=2 textures on the same wall).
I just tried drawing a new wall (with no bottom texture swap on this new wall), and then pasting the glitchy texture onto it. So, it appears glitchy in 32bit, as expected. Then, I pasted a non glitchy version of the texture on this new wall, and that fixed it for this wall.
However, if I do the bottom texture swap bit, it can no longer be fixed this way.

(then I thought perhaps the issue was dependant to the top texture in those cases? But apparently not.)

So in other words, I have found a solution to my problem : I'll have to waste tons of walls and sectors and add tiny new sectors everywhere so that the top and bottom textures are different walls, like this no need for bottom texture swap and I can paste the non glitchy textures. (Because, yes, the previous solution, using another texture, didn't satisfy me because that would be sacrificing the maps looks even in 8bit)
That's how much effort I'm willing to make the map look not glitchy for all the tasteless and clueless people who play Duke3D in 32bit and that the said 32bit makes things look glitchy (Yes, I'm pissed, I had to let this out). You know, I think that after that I'll only care about my how my maps look and play in dos Duke, everything else is just a painfull mess.

[Micky C]

It's because the original mappers compensated for the issue by shifting the textures in their maps. The bug's still there, they just covered it up.

[MetHy]

Mickey C, on 26 May 2014 - 03:30 AM, said:

It's because the original mappers compensated for the issue by shifting the textures in their maps. The bug's still there, they just covered it up.

That's assuming that the issue do come from the fact that the texture isn't power of two; by if that was the case I'm inclined to think ALL the use of that textures would look glitchy; except only about half do (like I said, the same non power of two texture used in the same manner, only about half look glitchy).

[Fox]

Not sure what you meant. Isn't it because half the textures are bottom-aligned?

[MetHy]

Check the example I joined here
http://forums.duke4....post__p__196956

They're all bottom aligned, but only half glitch.

[Fox]

The left and right sides have different y-pannings, 107 and 11 respectively. Which is a problem of non-power of two textures (107 - 11 = 96 = 2^4 * 6).

[MetHy]

Yeah I noticed that, but of course it you set the same panning it won't look aligned. And that still doesn't change the fact that alignment will be different in 32bit, no matter the panning, but only for the part on the left.

[Daedolon]

MetHy, on 26 May 2014 - 02:58 AM, said:

Edit : it seems that the reason why I can't fix the glitchy textures simply by pasting the same non glitchy texture onto it, is because in this case the glichty textures are at the bottom of a texture swap bit (=2 textures on the same wall).

The way I get around this is just change the texture from top/bottom to the other one. Sometimes moving the sector just by a notch vertically helps too (alt and mouse or just changing the Z in 2D mode).


EDIT: So yeah, I took a look at the map and the Cstat for the walls with the bug were different from the working ones, looks like they were flipped. I'm thinking since it's not power-by-two, the wall still gets shifted either way in 8bit because of the rendering issue. I just think due to sheer luck it ended up being shifted to the correct place on the wall to the right, but because of the vertical flipping the wall on the left had, it got shifted in a different way and ended up being rendered wrong always in one of the renderers.

[OpenMaw]

Does showview still not work in Polymer?

[Daedolon]

Nope. Waiting for it to be fixed so one of our mods can be actually released.

[MetHy]

Daedolon, on 26 May 2014 - 07:07 AM, said:

The way I get around this is just change the texture from top/bottom to the other one. Sometimes moving the sector just by a notch vertically helps too (alt and mouse or just changing the Z in 2D mode).


EDIT: So yeah, I took a look at the map and the Cstat for the walls with the bug were different from the working ones, looks like they were flipped. I'm thinking since it's not power-by-two, the wall still gets shifted either way in 8bit because of the rendering issue. I just think due to sheer luck it ended up being shifted to the correct place on the wall to the right, but because of the vertical flipping the wall on the left had, it got shifted in a different way and ended up being rendered wrong always in one of the renderers.

I'm not certain the flipping is related to the glitch. I've seen other textures glitch while they weren't flipped and while the same texture, same flipping, just next to it, didn't glitch. I also don't believe it's by luck that it ended up being correct; in the map there were a few of those all the around the map which didn't glitch, while all the glitchy ones glitched in the same way.

It's the 4th time I see this glitch, to me it's completely random. Thanks for finding a way to get around the problem though.

[Daedolon]

Daedolon, on 26 May 2014 - 07:07 AM, said:

EDIT: So yeah, I took a look at the map and the Cstat for the walls with the bug were different from the working ones, looks like they were flipped.

Helixhorned: Basically what this means, is that when pasting textures (tab->enter) on the lower portion of a two-textured wall segment (created by pressing 2), the cstat (slash flip) information isn't pasted among other info.

I'd count this as a bug.

[MetHy]

So I just realized that wall-aligned sprites aren't affected by light changes made with SE8 (up open door lights). Only when they're wall-aligned though, it works for any other type of sprites (like, it works if that same sprite is floor aligned etc); also, wall aligned sprites DO get affected by light changes using SE12 (light changes using a switch).

So, could this be something that could be implemented in eduke32? I don't believe it would influence already existing maps (wether from the original game or usermaps) in a negative way. If anything, I don't think wall aligned sprites and SE8 are used together very often or else somebody would have noticed this sooner...

Rather than being automatic though, perhaps it could be added as a feature the mapper has to 'turn on' by pressing a key on the sprite. I know that would make things more complicated but at least we'd be sure it wouldn't have unwanted consequences on already existing maps...

[Helixhorned]

MetHy, on 30 May 2014 - 06:29 AM, said:

So I just realized that wall-aligned sprites aren't affected by light changes made with SE8 (up open door lights). Only when they're wall-aligned though, it works for any other type of sprites (like, it works if that same sprite is floor aligned etc); also, wall aligned sprites DO get affected by light changes using SE12 (light changes using a switch).

That they're not affected by the shade change has nothing to do with SE8, but with the fact that wall-aligned sprites in particular don't take over the ceiling/floor shade. The code path responsible for this is in G_DoSpriteAnimations(). Note that this means that it's tsprite shade, not sprite shade:

    for (j=spritesortcnt-1; j>=0; j--)  // Loop over drawn sprites
    {
        spritetype *const t = &tsprite[j];  // 't' is a pointer to a drawn sprite
(...)
        if (t->picnum < GREENSLIME || t->picnum > GREENSLIME+7)
            switch (DYNAMICTILEMAP(t->picnum))
            {
(...)
            default:
                // NOTE: wall-aligned sprites will never take on ceiling/floor shade...
                if ((t->cstat&16) || (A_CheckEnemySprite(t) && t->extra > 0) || t->statnum == STAT_PLAYER)
                    continue;
            }

        // ... since this is not reached:
(Code that potentially takes over a ceiling/floor shade to 't'...)
    }

In contrast, the SE12 handling code explicitly sets the shades of wall-aligned (and only those) sprites contained in sectors affected by it. In G_MoveEffectors():

        case SE_12_LIGHT_SWITCH:
            if (t[0] == 3 || t[3] == 1)   //Lights going off
            {
(...)
                for (SPRITES_OF_SECT(SECT, j))
                    if (sprite[j].cstat&16 && A_CheckSpriteFlags(j,SFLAG_NOSHADE) == 0)
                    {
                        if (sc->ceilingstat&1)
                            sprite[j].shade = sc->ceilingshade;
                        else sprite[j].shade = sc->floorshade;
                    }
(...)
            }

^ That potential taking over of shade + SFLAG_NOSHADE check is repeated a couple of times throughout actors.c, but in slightly different variations. Yuck!

Quote

So, could this be something that could be implemented in eduke32? (...)

Rather than being automatic though, perhaps it could be added as a feature the mapper has to 'turn on' by pressing a key on the sprite. I know that would make things more complicated but at least we'd be sure it wouldn't have unwanted consequences on already existing maps...

In principle it could be implemented as another cstat bit. The problem is, this is really ugly... we'd be assigning meaning to a bit of an engine-side structure (sprite[]) in order to work around deficiencies in game code. Maybe these kind of things are good candidates for a (not yet implemented or even thought out) feature of the map-text format to carry ancillary data for map constituents.

[Helixhorned]

Daedolon, on 26 May 2014 - 06:20 PM, said:

Helixhorned: Basically what this means, is that when pasting textures (tab->enter) on the lower portion of a two-textured wall segment (created by pressing 2), the cstat (slash flip) information isn't pasted among other info.

I'd count this as a bug.

Yup, kind of. With r4493:

Quote

Mapster32: when pasting onto the bottom part of a [2]-wall, take over y-flipping.

Note that x-flipping is determined by the cstat of the upper part of the wall
(that is, the wall facing the player, not the nextwall, from which the picnum
for the bottom part is taken.)

So, if we were to take over x-flipping, it would affect the top part too.

[Micky C]

The mapster32 undo feature skips wwwwwwwwwwwwwwwaaaaaaaaaaaaaaaaayyyyyyyyyyyyyyyyyyyyyyyy too many changes involving things like pasting textures and altering texture alignments. Very annoying when you use the feature and have to redo a minute's work every time you wanted to undo one little thing. Is this a bug or an attempt at convenience by skipping potential insignificant changes? It's also a bit ironic because there was some other type of change that takes way too many undos to actually properly undo, but I can't remember what it was.

[Daedolon]

It does? I used it once in my map over a really small thing and saved, now I can't remember what I fucked up

[Hendricks266]

The first WIP of my new menus are committed. This was a task that took me well over 100 man-hours and was 11 years overdue. Besides cleaning up "the single worst fucking mess in all of EDuke32", it also opens up the door to custom menus down the road using Lunatic.

This has nothing to do with scripting for the time being, or CON likely ever.

The main impetus for this was making the menus Android-ready. Otherwise, there are not many changes that are visible but I'll point out the improvements I did make.

First off, all menus will automatically scroll if their contents exceed the page. You can see this with the Game Setup page, which is two screens merged into one, and the Keyboard Key and Joystick Button menus.



Slider bars will generally display a numerical value to their side. (I added a separate music on/off.)



You can press Enter on some menu option toggles to bring up a menu of possible choices. This was initially created for the mouse/joystick button menus, but I thought it natural to extend the functionality to places that could benefit.



In addition to this, you can press left/right on options you may not have been able to before, such as mouse/joystick input screens.

Smoothed out/improved joystick menu, though the underlying functionality is still broken and needs to be addressed.



One final thing is that you can enter the color escape sequences in a visible manner (like markup) while entering a string, then they will take effect when you exit the editing mode.



I still have yet to implement mouse/touch/pointer support, which is its own can of worms due to gesturing and whatnot.

Please test this thoroughly, and report any crashes or misbehavior.

[Helixhorned]

Hendricks266, on 31 May 2014 - 04:37 AM, said:

The first WIP of my new menus are committed. This was a task that took me well over 100 man-hours and was 11 years overdue. Besides cleaning up "the single worst fucking mess in all of EDuke32", it also opens up the door to custom menus down the road using Lunatic.

An enormous effort indeed, but one that will hopefully pay off in the long run. I guess reviewing it will take a while (git --stat shows 8334 changed lines in menus.c!). If I recall properly, you did all of your development in one file, and then kind of merged that in with menus.c, right? I'm asking because it would have been a bit easier to see commits broken down a bit more. But I can understand if that was too cumbersome to pull off given that this was to a large extent coding with a chainsaw.

Right now, it's failing to compile for me:

$ LTO=0 WITHOUT_GTK=1 RELEASE=0 LUNATIC=0 make -j4
(...)
source/menus.c: In function ‘M_MenuEntryOptionModify’:
source/menus.c:2475:24: error: ‘ME_GAMESETUP_UPDATES’ undeclared (first use in this function)
     else if (entry == &ME_GAMESETUP_UPDATES)
                        ^
source/menus.c:2475:24: note: each undeclared identifier is reported only once for each function it appears in

Edit: OK, that one only needed an #ifdef _WIN32. Looks good.

[Fox]

Hendricks266, on 31 May 2014 - 04:37 AM, said:

Would it be feasible to classify it by aspect ratio?

Aesthetically, the text should be in the left corner, or center-aligned, or being aligned with the x in the center.

[Daedolon]

I'd like them to be sorted out by height > width and having the largest resolution at the top, with all the text aligned to the right.

[Helixhorned]

MetHy, on 24 May 2014 - 01:20 PM, said:

I've already mentioned this glitch in another thread and I found a way to 'fix" the problem but the 'fix' I found doesn't work anymore.

Here is the thing, some textures get misaligned when switching to 32bit mode.

That's a well-known issue, its root cause being the improper rendering of walls with non-power-of-two height textures in 8-bit mode. Here's an example with BIGORBIT2 which is h := 400 texels high:


See what classic does here. Basically, it renders the wall as if with a texture of a height that's the next power of two greater than h, 512. This pseudo-texture has 400 horizontal lines of the original texture, plus 112 more, made up from the first ones of the same texture. As a result, there is now a seam in the apparent texture. What's the problem with that? Well, the Polymodes have no problems with NPOT height tiles at all:


But this means that such walls can't always be displayed the same in both renderers. A necessary but not sufficient condition is that a wall can only show at most h texels in height.
In order to overcome certain improperly aligned cases in the original maps, Polymost, and then later Polymer added a hack that modifies the apparent ypanning of a wall when its real .ypanning exceeds some value (that's also dependent on h). However, this workaround suffers from the obvious drawback of introducing a discontinuity with ypanning.

Enter r_npotwallmode (r4498):

Quote

Polymost: r_npotwallmode, emulating 8-bit for walls w/ nonpow2 height textures.

When that mode is enabled (see below for caveats), wall textures that have a
non-power-of-two height (call it 'h') will be modified to look like in classic:
Let 'H' be the next power of two greater than 'h'. The texture will be uploaded
with height 'H', made up from 'h' hlines of the original texture, followed by
'H'-'h' first hlines of the same.
No panning "corrections" will take place. The mode is disabled by default.

Caveats/notes:
* the mode requires that r_hightile is disabled
* it is not implemented in Polymer
* in the Lunatic build, it is ineffective when a VX map is loaded, as those
display walls with NPOT height textures correctly

[Jblade]

Anyway to disable the darkening effect of the menu on the gameworld?

[Helixhorned]

James, on 01 June 2014 - 08:30 AM, said:

Anyway to disable the darkening effect of the menu on the gameworld?

Currently no, but feel free to add it to the suggestions wiki page under a new section.

[Fox]

Can you add a function to offset the menu up or down (at least for the main menu)? This is an issue if you are using an header tile bigger than the original.

[LeoD]

Synthesis builds are broken - last version created is r4488.