Possible slicing bug?

[errolt]

Hi everybody,

I don't know if I'm being silly or broke a setting somewhere but I'm having strange problems slicing a part from http://www.thingiverse.com/thing:41967

I've run the part through netfabb basic which finds no problems.

This funny thing is that I have two identical parts on the build tray, but they are sliced differently.

Also, exiting preview mode and re-slicing produces different results every time...

I also don't know why there are gaps in the outside perimeter...

Slice 1

Slice 1

Slice 2

Slice 2

Thank you,
Errol

[3Drasle]

Yes, I have the same problem.
But can all ways get it to slice correct if I change the layer hight a bit.

/Michael

[Frguy]

I have approximately the same problem

Sometimes bad layers appears (already present in 2.0.1 v).
The bug disappears by slightly modifying the height of the first layer (or all layers).
I think to a calculation problem when the layer has a critical height /certain forms.

Here are two examples of the one part. The bad layers appear at the beginning and at the end of a horizontal hole.
I printed this part without any defect after modification of the first layer, but it is important to check slicing before printing.

[errolt]

Wow, can't believe it. Changed first layer from 100% to 101% and now it slices fine.

[zakkus]

yeah i have this one too (with 2.1.0). i cant seem to get it to go away, no matter how much i fiddle with the layer settings.

[jimc]

zak have you run the model through netfabb to make sure its ok. many time when this happens its a model problem but many times its a problem so slight that programs like netfabb dont pick it up either and it just makes s3d react funny when it shouldnt really be anywhere near that sensitive. hopefully they get this worked out next release

[zakkus]

yeah, that seemed to be it. I am hardly a netfabb expert, but it seemed to fix "something", and the weird tessellations disappeared.

[dalew8abz]

I suspect something related to doing "exact" equality comparisons of "inexact" floating-point numbers, like testing if someVariable >= 1.0 when someVariable is the result of a calculation that produces, say, 0.999999999998 or some such.

There are floating-point math libraries that can do "near" comparisons instead of "exact" comparisons (with a programmer-specified tolerance) to handle this sort of situation.