ghost pixels with "Use global LED buffer"

[softhack007]

What happened?

The "Use global LED buffer" (LED preferences) option can add ghost pixels which are not there without global buffer.

To Reproduce Bug

It seems that "ingredients" are

• using global leds buffer, plus
• some segments have width different from the overall 2D layout, plus
• a smaller segment overlaps with the bigger one, plus
• a) the bigger segment does something that reads buffered leds -> blur, fade-out, move, ...
• b) or the smaller segment is an overlay

Example:

• global leds size 24x37
• segment 0: black hole effect
• segment 1: android effect
• segment 2: size 12x8 (scrolling text overlay "45")
• the white pixels to the right of "45" should not be there. They are "ghosts" from segment 2.

First Analysis

• when drawing a pixel into segment[2], two things happen in setpixelcolorXY
• 1. if (ledsrgb) ledsrgb[XY(x,y)] = col; // buffer assigned to segment
• 2. strip.setPixelColorXY(start + xX, startY + yY, col); // write-thru to NPB driver

the "ghost pixels" are from the write to ledsrgb, which assumes wrong buffer coordinates; XY() is computed using segment boundaries only. As Segment#0 later uses "blur()", the global ledsrgb buffer pixels are used and ghost pixels appear next to the white text.

Conclusion

It seems we need a different XY() function for global buffer, as it must consider the "global" width instead of segment width. I think this new global_XY() needs to also use rotation and mirroring of a segment in the calculation.

Expected Behavior

Effects look (almost) the same, independent from using global buffer or not.

Install Method

Self-Compiled or other

What version/release of MM WLED?

any 0.14.0 build

Which microcontroller/board are you seeing the problem on?

ESP32

Relevant log/trace output

No response

Anything else?

This problem exists since "global buffer" was introduced in upstream WLED 0.14.
We can either

• wait until 0.14.0-beta4 is available (where global buffer will be removed)
• find our own solution for MM (in case we want to keep global double-buffering)

To avoid wrong pixels, I would suggest to not use global buffer unless all segments have the same width as your overall layout. Without global buffer, there are no wrong pixels.

Code of Conduct

• I agree to follow this project's Code of Conduct