Even after fixing the regression of r2564 this bug persists. Some simple logging shows the remaining issue introduced in r2591:
Here I swipe 4 fingers and release them all:
4 touches, 0 released
4 touches, 1 released
3 touches, 0 released
3 touches, 0 released
3 touches, 1 released
2 touches, 0 released
2 touches, 1 released
1 touches, 0 released
See how Mir thinks one finger was never released. That's the bug.
Even after fixing the regression of r2564 this bug persists. Some simple logging shows the remaining issue introduced in r2591:
Here I swipe 4 fingers and release them all:
4 touches, 0 released
4 touches, 1 released
3 touches, 0 released
3 touches, 0 released
3 touches, 1 released
2 touches, 0 released
2 touches, 1 released
1 touches, 0 released
See how Mir thinks one finger was never released. That's the bug.