It looks like Arcanist included more changes in the diff than I intended. I have updated the diff manually.

Okay, I have pushed the patch. (I still can't figure out how to close the revision.)

I must be doing something wrong with my commit messages. I included "Differential Revision: https://phabricator.kde.org/D6263" in the merge commit message and pushed, but it doesn't seem to have closed the revision. I don't see an option to close the revision through the Web interface either. Do you have any ideas as to what I am missing?

Okay, great! I will make the changes you suggested and merge to master.

Okay, great! I will change those asserts then merge into master. I still want to do something about the issue of having to wait 1 second if the pressure curve goes to zero, but I will work on that in a separate revision.

@dkazakov: I think this new diff addresses the issues you pointed out in the previous diff. I have added back the safety check in KisToolFreehandHelper::doAirbrushing, removed the reliance on parsing string options during painting, and added the call to qBound you suggested. I have also changed the KisToolFreehandHelper::paintAt workaround so that the initial paintAt call only happens when airbrushing, which should avoid the zero-rotation dab at the beginning of a stroke when angle-dependent brushes are used without airbrushing. For the case where an angle-dependent brush is used with airbrushing, I added the ability for KisToolFreehandHelper to supply an initial drawing angle based on the most recent cursor movement, so the initial dab's rotation will at least make some sense. I also changed the angle-computing code so that, if two consecutive dabs are at the same position, the drawing angle from the previous dab is reused. Airbrushing and angle-dependent brushes still don't work very well together, since it can be hard to avoid small movements that change the angle when trying to airbrush at a single point, but the behavior seems at least somewhat reasonable now.

I don't think it was. Should I merge it myself? Is there any particular protocol to follow with merging revisions to master?

Thanks for the feedback. I will try to get the issues fixed soon.

I have made a new version of the diff that addresses some of the issues, but I think further discussion is needed. In the new version, KisDistanceInformation clamps the airbrush interval to the range [0.5, 1000.0], meaning the rate can't be less than 1 dab per second. I changed the code back to only updating the spacing when a dab is painted, which should fix the problems I introduced with the previous diff. It is still possible to get an unpleasant effect where the pressure can make the dab rate slow (but no slower than 1 dab/s), forcing the user to wait 1 second before the next dab. This especially happens at the start of strokes, as the pressure is low when the pen first touches the tablet. The ideas I have for solving this are:

1. Increase the minimum rate to more than 1 dab per second, so that it won't be as noticeable.
2. Implement some form of spacing updates that can happen between dabs. From my previous attempt at this, it seems prone to causing bugs and performance issues.
3. Leave it as is and let users work around it by setting a Rate curve that doesn't go too low. (Possibly changing the default Rate curve as well?)

We need to discuss which idea works best, or if there are any other solutions.