Created attachment 162704 [details] Screenshot of wiggles I've set up a 2-point-assistant and a vanishing point. If I draw a line with snap to assistant that's supposed to snap to the 2-point assistant, the start of the line often has a small wiggle. "Snap to single line" is activated. See screenshot. Drawing the same line in the reverse direction usually doesn't get a wiggle. Snapping to the single vanishing point doesn't produce any wiggles. If I remove the single vanishing point, no wiggles either. If I go back to Krita 5.1.5, no wiggles either with the same setup of assistants.
Created attachment 162705 [details] Example kra file with assistants set up
Playing around with it further, I think the line wants to snap to the single vanishing point first before it decides to latch onto the 2-point assistant. Only that I have "Snap to single line" enabled where it should stick with the line first snapped onto. The less precise I am with my line, the bigger the wiggle. For some lines, it's almost impossible to be precise enough to not get a wiggle at all.
Another user has reported the same behaviour on KA: https://krita-artists.org/t/problem-with-combining-2-points-perspective-vanishing-point-assistants-in-version-5-2-1/77984/2 I've further found it's not only the 2-point perspective assistant that's causing this, just the one most likely/frequent to behave like this.
Seems I didn't properly look for bug reports, I wasn't even aware it regressed from 5.1.5. Can you test with current master? The old setup for nightly builds is gone unfortunately, you currently have to click your way through https://invent.kde.org/graphics/krita/ to download a CI "pipeline artifact" of the last successful build (that hopefully hasn't been deleted yet). If no one finds regression compared to 5.1, we need to backport https://invent.kde.org/graphics/krita/-/merge_requests/2066 to 5.2 I guess.
Yeah, the wiggles are gone now. I tested with the file I had attached to this bug report, and also another file where I tried out my most frequent use cases. Overall, it works well for me now.
Thanks for testing. Backported to 5.2 branch, so it should be in 5.2.3 (commits b70eaea4 to 5db332f8)