LDCad 1.7 Alpha 1 (win+linux)

[N. W. Perry]

You mean it goes wrong when a submodel also uses buffer exchange or the whole (non exchange) model is counted?

I meant the second, but it may actually be both in this case.

Could you send me the problem file?

Yeah, here it is. (Note the main subfile is currently empty 'cos it's not finished yet.)

This model is actually progressively recursive, but I don't think that's the issue as it's present even at the first level of nesting. What I mean is that the submodel for box 1 is a subfile in box 2. Then that box 2 submodel is a subfile in box 3, and so on. So the double-counting starts to add up pretty quick.

I also include per-step part counts in the comments, which also reflect things like u-joints being actually one part instead of three, not counting stickers as parts, etc. So that doesn't exactly match LDCad's count, but I can at least account for the discrepancies.

(By the way, this is the same model in which I was having the snapping issue with axlehole prims, so maybe it will help to recreate that.)