Heightfields with negative values will fire face normal errors

[rogerscg]

To reproduce, change all heightfield data points in the heightfield.html demo to -5. You will see 1,000+ errors thrown from ConvexPolyhedron when computing convex triangle pillars.

This was also an issue from 2014: schteppe#149

[rogerscg]

The failure to collide might just be due to lack of CCD, not a heightfield issue. Removing that part of the issue (though still a big issue).

[rogerscg]

I found the issue, which stems from the convex triangle pillar being upside down when the heightfield has a negative value. Will send a PR in the next few minutes.

[codynova]

Awesome, great work digging in and hunting down the issue. I'll cut a quick release with your fix.

[codynova]

I've published a new version of cannon-es with the fix to heightfields cannon-es@0.15.1

[rogerscg]

Awesome, thanks for being so responsive.

[codynova]

Absolutely, we welcome any further contributions 👍