Just spotted this. Pretty poor regression.

This was changed in #7290.

Before considering this a regression we want to fix (a negative width is arguably undefined behaviour?), @martinleopold and @ff6347 do you have opinions on what the expected behaviour here should be?

@davepagurek upon double checking, I'm not sure that is where the change happened, as the changes in that PR is for arcs. however, comparing 1.11.0 (which works) and 1.11.1, i can't see anywhere else that might of caused this regression/change

I believe modeAdjust, which was modified, is also is used by rect, and they added unit tests for rectangles too.

Also ccing the rendering stewards in case anyone has thoughts: @limzykenneth, @ChihYungChang, @teragramgius, @tuminzee, @Zarkv, @robin-haxx, @Gaurav-1306

This was changed in #7290.

Before considering this a regression we want to fix (a negative width is arguably undefined behaviour?), @martinleopold and @ff6347 do you have opinions on what the expected behaviour here should be?

I think this was caused by my PR. It's pretty much undefined behaviour as the reference doesn't talk about negative widths and heights. Some comments in the code were mentioning "p5 supports negative width and heights for rect" but not more.

Note that prior to my PR there were virtually no tests for the basic shape rendering/and shape modes, so no 'defined' behaviour in that regard.

Using negative width/height to flip rectangle/ellipse/arc does make sense to me though – for rectMode(CORNER) at least. I think we should revert to this behaviour and use the opportunity to update the docs accordingly as well.

Negative height and width has been part of p5js for as long as I can remember. This is certainly an accidental regression and a breaking change. It makes 1000% sense as it's the easiest way to draw rectangles from the bottom up due to p5's "odd" coordinate system

Yes, it is a regression in this case. Feel free to file a PR for a fix.

I just ran into this too – I usually render FFT audio bars with something like rect(x, height, w, -freq), and just wondered why it stopped working?! Indeed just like circle() and ellipse() the negative value has always been an inversely drawn form since many versions ago.