At first, I was thinking this was a duplicate of #5875 (which is fixed in master), but I see that this is different. The problem here is that ax.plot() probably isn't properly normalizing all of the various keyword arguments, and as such doesn't detect that the cycler and the keyword property are colliding.

The plot() function explicitly normalizes the 'c' keyword into 'color', but it doesn't do it for any other properties. I think I can put together a patch for this utilizing the same normalization tool that the cycler goes through. @tacaswell, wasn't there another normalization mechanism that was added recently?

https://github.com/matplotlib/matplotlib/blob/master/lib/matplotlib/cbook.py#L2449 but we need to use it places now.

but we need to use it places now

Sorry... my parser couldn't grok that sentence.

I mean we have the function available to use, but we need to actually use it in plot, etc

Ah. Yeah, I started looking into that, and I am not entirely sure it is ready for prime-time. At least, we don't really have the supporting data for it. One of the big problems I see is that it requires me to know what other keyword arguments will be allowed that don't go through the aliasing mechanism.

Perhaps it makes sense to break this function out into three levels. The first level simply does normalization (like I do in the validation step of prop_cycle), and the second level does precedence handling (conflict resolution), and the third level does the "required"/"allowed" handling?

the allowed parameter only does something if it is not None, otherwise it lets things pass through.

I created #6345 that should address this problem. Note that v1.5.x will still have a problem if your manually created property cycle uses aliased property names, but that is fixed in 2.x as I noted earlier.

This seems to solve the problem properly so I closed the issue.

Thanks for the quick response!