Thanks for the PR - are you still interested on working on this? Before it's ready to be reviewed, it still needs a test adding to make sure we don't accidentally introduce the bug again later down the line.

Thank you for the comment! I am interested to put more work in, but not yet sure how to go about tests. The bug is caused by code in the indicator's draw() method, specifically these lines:

Because of that the result of get_clip_on() is irrelevant to the final output. Do you think an image comparison test (doc below) is the way to go or is there a better way to test what is happening in draw()?

https://matplotlib.org/devdocs/devel/testing.html#compare-two-methods-of-creating-an-image

The easiest way to add a test would be to modify this existing test:

by setting the clip_on value of one of the connectors at the end. If all works well, the new test image show show that one of the connectors is clipped.

Hi @juseg are you still interested in working on this? Do you need any more help?

Hi @rcomer thanks for the heads-up and sorry for letting this sleep, I am back at it will try and add tests based on the advice above!

• I reworked the image comparison test as suggested.
• Amended another test on rectangle and connector properties.
• I added an example using clip_on=True on the rectangle but not the connectors to draw an inset indicator over broken axes (this is actually my original use case for wanting this implemented).

I am happy to get feedback at this stage. Specifically:

• Is this considered "new features and API changes" (see PR checklist)?
• Do we want clip_on=True by default on the rectangle, while keeping it False on the connectors (it would seem like a logical choice to me, but also a breaking change)?

Edit: @melissawm Github Actions seems to remove that topic: axes tag with every commit.

Huh! Looks like a bug in our labeler action - I'll take a look. Thanks for flagging @juseg !