@HooglyBoogly the assert in sample_curve_attribute fails because it assumes the src and dst geometry has the same number of curves, when in fact only the index mask sizes have to match. This…
My impression was that we add such system includes at the bottom, although checking other files it seems to be quite arbitrary. There doesn't seem to be anything in the style guide about that…
I've tried to avoid stuffing everything into a single giant file, although the boundary is somewhat arbitrary. None of these things need to be declared here since the fill tool is the only using…
This is the best i could come up with so far. DisjointSet doesn't actually help with anything, since we have to iterate over the curves anyway. This also does not preserve existing order of…
@filedescriptor Yes, as far as i can tell the newer library has the same issue (maybe i missed something while testing though).
Allocate another one array, and for each group, write index in a loop like i != -1; i = next[i] { dst[group[i]] = i; }
That would be O(n^2) if i'm not mistaken. That's probably (?) fine for…
Nice idea, but unfortunately knowing the sets of curves going together is not enough: i need to actually re-order them so that the points form a contiguous range. The AtomicDisjointSet does…
It's starting out with uninitialized dst arrays in the mix_arrays function. There needs to be some zero-initialization before adding in values.