I cannot replicate:

Task T68943 was related to this too.

It looks there is some type of Overflow in the UI code when the range is near of the INT max values. I have replaced by SHRT_MAX because is enough.

I can confirm with daily build.

I have tested these 3 arrows and works for me.

This solves T68876.

I just tested the las code compiled and I cannot reproduce it.

By design, the array works by Layer, so each layer is created one by one for all array elements. As the drawing order is defined by Layer position, you can get this problems. To fix it, just put all drawing in one layer only (this is required to get good result with array).

@Brecht Van Lommel (brecht) any update here? I need implement this patch before working in the Draw Engine (T66293) and other changes.

@Aaron Carlisle (Blendify) Maybe it would be good add a note. When you use only selected points (default) the merged points must be contiguous. I mean, if you have unselected points betweeen selected, they will not be merged.

Please, @Matias Mendiola (mendio) and @Daniel Martinez Lara (pepeland) verify now it's impossible sculpt not selected points when mask was enabled, I have detected a problem in that code and fixed.

• GPencil: Fix wrong mask flag checking for limit Sculpt

@S J Bennett (quollism) You can close the revision. Thanks for your help fixing this bug.

@S J Bennett (quollism) Do you have commit rights?

For me it's ok, jsut replace the comments before commit the patch.

I updated the diff, and ARC created an update of this revision, instead of update it (D5504). I keep this revision because the changes were small and I will commit manually, not applying this revisión, so the final commit will be slightly different (small cleanup).

Created by error using ARC (real is D5500)

• GPencil: Remove old code not used
• GPencil: Remove mask button for Sculpt and use Select Modes
• GPencil: More work to decouple Select modes

Proposed fix: D5500

Proposed Fix: D5489

Reduce pop whith extreme factor values.

This is not a bug. Since 2.80, the add-on generates the measures in a separated image. To get a final image, you must use compositor as this example:

Thanks for the code, this solves the problem, but I'm thinking that maybe would be better Normalize the points in the vertex group. I mean, if only part of the stroke is in the Vertex Group, then normalize only those points.

Not sure if the naming would be better with the changes above.

• GPencil: More cleanup to reduce patch size

I moved the struct definition to old place to make patch small.

• Cleanup: Move struct definition

• GPencil: Rename variables adding gpencil_ prefix
• GPencil: Move code outside BKE module
• GPencil: Cleanup Marcro names adding GPENCIL prefix

@Jacques Lucke (JacquesLucke) Do you agree with the proposed fix?

Also, thinking is how GP data works, as you have different information in each frame, to evaluate all datablock, I would loop all frames, but if I calculate only the current frame, then make full copy of the datablock is not logic because we are coping something we are not going to use.

I know the mesh use a derived mesh datablock, but there are reasons to do in this way:

Proposed fix (now when you enter in any mode, the object is always dsiplayed in Dopesheet):

There is one point here. The Dopesheet is reused for Annotations and this is not related to Active Object, so we need be careful about that.

New Patch here D5470. This patch is only yhe depsgraph code.