Problem with addon RoundedBoxGeometry in version r178

[jpp]

Hello

Today I downloaded the most recent version (r178) of threeJS and I found a problem with addon class RoundedBoxGeometry which didn’t show up in previous versions. One can see it here (look at the tiles of the puzzle):

Sliding Puzzle

I compared files RoundedBoxGeometry.js in versions r178 and r177. Based on that, I realized that if line 73 in r178 file:

super( width, height, depth, segments, segments, segments );

is replaced by this one:

super( 1, 1, 1, segments, segments, segments );

the problem gets solved.

So, I kindly ask if the above-mentioned problem is due to some bug in RoundedBoxGeometry.js or if I need to make some changes in the code of the puzzle in order to adapt it to this (and to future) versions of threeJS.

Thanks in advance.

[manthrax]

The puzzle isn’t loading for me? Just getting a white screen…

no errors in the console.

[manthrax]

Oh ok it finally loaded…
But upside down:

[jpp]

4x4 and upper puzzles load upside-down and with a awful look.

2 x 2 and 3 x 3 puzzles don’t load upside-down, but they look awful as well.

Now please try running it with version r177. You will notice that it will load and look fine.

[manthrax]

can you set material.wireframe = true and post screenshots of the 2 versions ?

I don’t know how to run/switch your versions.

[jpp]

Will do so in a moment. Switching from mobile to notebook…

[jpp]

material.wireframe = true (r178):

material.wireframe = true (r177):

It should look like this (with material.wireframe = false):

It looks fine with version r177 and, I as wrote before, if in line 73 of file RoundedBoxGeometry.js one replaces super( width, height, depth, etc. ); with super ( 1, 1, 1, etc. ); (similar to previous versions of this class).

[manthrax]

@Mugen87 is this a regression?

(related ?): RoundedBoxGeometry: add type, parameters and toJSON (#31340) · mrdoob/three.js@6b2f433 · GitHub

Looks like the super( constructor originally took 1,1,1 and then later roundrect deformed the vertices.. but that change is passing width/height/depth to the constructor instead?

[Mugen87]

Yes, this was a regression and is now fixed on dev. The next release r179 should restore the previous behavior.

[manthrax]

thanks!!