Actually, the reason for bug 877019 is simple: it's difference is size! The sploth-bad object is just small, and on this scale, the offset algorithm is erratic. When applied to a larger object, it works much better. Can you please have a look at what's wrong with this?

...

• It appears that all offsets now use the imprecise

algorithm (you once mentioned that there were two algorithms, one precise but slow and the othe imprecise but fast). I think we should use the precise algorithm for all kinds of offsets, dynamic or not.

the "imprecise" algo is now the more precise one; the problem whith the old "precise" is that sometimes it doesn't produce the mathematically correct offset, but is faster. this can happen when offseting the 'D' letter with a big radius, for example: a hole appears in the offset. since speed is an issue, the fast one is back in place.

also the livarot stuff now uses NR::Point (and some NR::matrix); it may contain some bugs since the conversion touched a big portion of the code. you can report any strange behavior in boolean operations and offset to me.

_________________________________________________________________ Help STOP SPAM with the new MSN 8 and get 2 months FREE* http://join.msn.com/?page=dept/bcomm&pgmarket=en-ca&RU=http%3a%2f%2f...