Layering issues in Puyo Puyo Fever 2 (Windows x64 Build)

[internetakias]

In PPF2, the layering is messed up when the renderer is set to OpenGL. Resulting in all the game objects besides the background being invisible. That is, util you wait long enough in the title screen for the demo puyo match to begin, then the game fixes itself for some reason.
Here are the display lists, one taken in the opening before a match and one taken after
https://gist.github.com/internetakias/6471825
As for screenshots, you can refer to this thread(Despite the big gap between the current version and the version used to take the screens, the bug remains the same): http://forums.ppsspp.org/showthread.php?tid=3858

[internetakias]

Any update on this? This issue has been around for a while...

[hrydgard]

7 days? Come on.

[internetakias]

No, I mean the layering issue, not this particular issue report

[internetakias]

Any info on this? The game no longer crashes at the opening so I think it might be a good time for someone to look into this too

[unknownbrackets]

Some framebuffer bugs were recently fixed, has this changed at all?

-[Unknown]

[internetakias]

I'm afraid not. The issue itself isn't related to buffered rendering anyway since it occurs regardless of whether it's on or off

[internetakias]

Are there any plans to look into this? It's kind of a big issue, considering you can't see anything but the backgrounds and all.

[thedax]

It's a depth issue like Phantasy Star Portable 2. If I use the same float depthRangeMin = zOff; "hack" as PSP2, it'll suddenly show correct looking graphics, but not 100% correct graphics (some intro/demo videos are still weird looking/missing things). There could be more to it than that, but that's one key element, anyway.

With the hack, I get a proper looking title screen:

Debugger shots without the hack:

[dbz400]

The most obvious fix for this hack is the logo below .Softgpu seems to be rendered all correct though

[dbz400]

If i uses this one , also renders correct

 float depthRangeMin = gstate.getDepthRangeMin() / 65535.0f;

[thedax]

Indeed, but it leaves PSP2 horribly crippled. Wonder what the true min and max are..

[dbz400]

Ooops this look better .Try a depth range check here .(Fix PPF2 and PSP1 , PSP2 remains unchanged)

    float min =  gstate.getDepthRangeMin() / 65535.0f;
    float max =  gstate.getDepthRangeMax() / 65535.0f;
    float depthRangeMin = zOff - zScale;
    float depthRangeMax = zOff + zScale;
    if (depthRangeMin > min)
        depthRangeMin = min;
    if (depthRangeMax > max)
        depthRangeMax = max;

EDIT :seems to be breaking Jeannac D Arc

[unknownbrackets]

Let me compare with softgpu just to explain:

float vpz1 = getFloat24(gstate.viewportz1);
float vpz2 = getFloat24(gstate.viewportz2);

float retz = coords.z * vpz1 / coords.w + vpz2;

coords.z / coords.w is the z value. If z is unsigned, then vpz2 is the minimum value. If z is signed, then vpz2 - 1.0f * vpz1 is the minimum value... either way, vpz2 + 1.0f * vpz1 should be the maximal z value (when z is from 0 - 1 as it is in GLES - this is not the case on the PSP or in softgpu.)

Now, looking at this code:

float zScale = getFloat24(gstate.viewportz1) / 65535.0f;
float zOff = getFloat24(gstate.viewportz2) / 65535.0f;
float depthRangeMin = zOff - zScale;
float depthRangeMax = zOff + zScale;
glstate.depthRange.set(depthRangeMin, depthRangeMax);

It sets the range to vpz2 - vpz1 - vpz2 + vpz1. This assumes that the depth is signed, which I don't know if is true or not. Removing the zScale part from the depthRangeMin equation should assume that depth is unsigned.

See my thoughts here about sign:
#4572 (comment)

Maybe it's depending on throughmode or something?

-[Unknown]

[dbz400]

Another try to reverse the position of zoff and zscale , it show up all the missing stuffs (but instead missing out other stuffs)

    float depthRangeMin = zScale - zOff;

[dbz400]

@unknownbrackets , but i think it is done in non-through mode already ? (though mode already set to glstate.depthRange.set(0.0f, 1.0f);)

[unknownbrackets]

Ah, right, and that's probably correct.

zScale - zOff is definitely not right.

It may even be a rounding issue.

-[Unknown]

[internetakias]

Like I mentioned above, when a puyo match begins, the game suddenly fixes itself. Do any of you know why that happens?

[unknownbrackets]

Has this improved at all with the recent depth changes?

-[Unknown]

[thedax]

No, it's still as broken as before. This game's depth problems are the same as PSP2's, I think.

[internetakias]

Not sure if this is even worth mentioning but the Dpeth range hack for PSP2 doesn't seem to work for this game.

[ppmeis]

Tested with latest build. Same status:

[daniel229]

Fixed in #7920