NPOT causes small visual glitch in naev

[ptitseb]

NPOT causes small visual glitch in naev

[ptitSeb]

Yes, LIBGL_NPOT=2 expose a function that is not completly supported for now. LIBGL_NPOT=1 is 100% emulated, but not 2 for now. I may work on that, but real support as not been mandatory for anything yet.

[MagaTailor]

I see, Expose limited NPOT extension sounded as if it might be too limited so I never tried that. My bad! :)

[ptitSeb]

Limited NPOT is : can use texture not NPOT, but cannot use Repeat or Mirror on NPOT, and no Mipmap neither.
basicaly what glshim does is can use NPOT texture (by creating a bigger NPOT texture), and so no Repeat or Mirror. Mipmap is supported but can lead to some artifact on the border (as you have already seen). That effect can probably be mitigated for textures with an alpha channel if I upload an empty texture first.

Full NPOT can do whatever they wants. Emulating that on non-NPOT hardware would mean resizing a non-NPOT texture to NPOT, meaning a lot of computational power at each upload, plus some artefact du to the resizing (espcially on SubImage manipulations).

[ptitSeb]

I'll close this ticket.

[MagaTailor]

Talking of glitches, just remembered this unrelated mouse pointer problem in TOME:

LIBGL_NOLUMALPHA makes no difference and capturing the screen normally shows just the regular X-window mouse pointer.

[ptitSeb]

did you retry with latest changes (with added Pixmap, PBuffer and multiple glX Context, it may have just solved the issue) ?
I never compile ToME, and didn't look at the code, so no idea how the mouse cursor is built.

[MagaTailor]

On Tue, 01 Nov 2016 15:25:14 -0700
ptitSeb [email protected] wrote:

did you retry with latest changes (with added Pixmap, PBuffer and multiple glX Context, it may have just solved the issue) ?

I was using the latest glshim master and it's still there. Any special options to try?

[ptitSeb]

Ah no. I need to look in the code to see how that cursor is built.

[MagaTailor]

On Tue, 01 Nov 2016 15:55:25 -0700
ptitSeb [email protected] wrote:

Ah no. I need to look in the code to see how that cursor is built.

Thanks!

[ptitSeb]

@petevine : it turned out it's not a glshim issue, but an SDL2 one. Check my github repo of SDL2, there is push about this: ptitSeb/SDL2@604d108

[MagaTailor]

Nice find! However, and I hope I'm not misremembering things, the white silhouette wasn't present on real hardware. Could the fix be incomplete?

[ptitSeb]

white silouette is there depending on the theme.

But if you want, you can change the if(alpha==0) by a if(alpha<25) to have something probably better (there are 2 way of putting cursor in X11 in SDL2, and the "legacy" way use something similar).

[MagaTailor]

Yes, it's much better that way but it still looks artifactishly wide. The white background is not visible on x86/OpenGL 1.5, just a thin blue outline. I fired up my old Athlon box to take this picture :)

[ptitSeb]

Not much can be done here. It's X11 functions, and it seems they behave differently for machine to machine.
Play with the alpha stuff, put alpha<125 to remove moste of the semi-transparent stuff, or even up to <255 to just discard everything not solid color (but that will probably remove the blue aura too)

[ptitSeb]

For the record, here is the original PNG image (for the dark theme)

[ptitSeb]

And here the image for default theme.