i didn't get the model of card, but 10.18.10.4358 was reported broken on irc by StarAndy (drawings show up delayed on the canvas)

in the processor part where it says "atom" it should say "ARM" instead.

behaviour when shrinking the docker ...

before:

This requires KisFixedPaintDevice to learn about LOD; that's a task that's slightly too big for me, as it probably involves a whole lot of stuff.

https://bugs.kde.org/show_bug.cgi?id=363089 also mentions this

At 20% zoom, drawing with a 1000px big airbrush_linear_noisy brush:

wolthera pointed out that @abrahams is currently working on the shortcut stuff, so maybe a CC is warranted

It seems like I will have to pull in another external dependency to get joystick/gamepad support.

I personally have a desire to be able to have different Tool Options set for eraser mode and non-eraser mode: When drawing, I like to have some sort of smoothing enabled, possibly the stabilizer with a strong-ish setting if i'm inking over a sketch for example. Unfortunately, having smoothing (or even worse: the stabilizer) turned on when erasing makes it very tedious to do the kind of back-and-forth motion I often use to erase. It'd be nice if the final design for this could accomodate my use case. Thanks!

landed the patch in master. the improvement will soon roll out to our users and i'm happy :)

@abrahams I've checked and the methods in question get called concurrently, so having the engines be shared would cause concurrent access to the seed again, which is what I had hoped to prevent with these patches in the first place. I don't know of a better approach than this, but if someone ever finds one, they are surely free to put it in :)

We've decided on IRC that KoColorspace should probably supply randomization functionality, so this patch is now, in my opinion, ready to land.

Dear @abrahams, with these new changes that use the c++11 random engines i still get the noticable speedup.

Use C++11 random instead of rand in fuzzy brushes and TotalRandomColorSource

i set boud as a subscriber, but not as a reviewer. my mistake!

For some reason the position tabletEvent->hiResGlobalPos() reports is relative to the top-left corner of the scratchpad widget.

boud asked to be subscribed via IRC

boud asked to be subscribed via IRC

boud asked to be subscribed via IRC.

the new diff version has been uploaded at the same spot.

this version of the diff uses the random number generator that's safe to use in concurrent situations to initialize a per-stack random state that then gets used in the inner loops.

actually, i was wrong: the class is used from multiple threads simultaneously, so the random state would get accessed concurrently. even though that's not problematic per se, as reproducibility is not needed here, i'm going to make a second version of my patch that creates a random state on the stack and seeds that from the global (locked) random number generator.