suspected mpl 3.4.1 regression for a graphical test

[bjlittle]

🐛 Bug Report

How To Reproduce

As part of #4087, the perceptual image hash images have been updated to account for the associated changes within iris from the recent release of matplotlib 3.4.1.

As part of #4087, https://github.com/SciTools/test-iris-imagehash/pull/39 was raised to update the test-iris-imagehash archive.

Consequently, there appears to be a regression for one the graphical test iris.tests.test_mapping.TestLimitedAreaCube.test_grid.0, see here for further details.

Steps to reproduce the behaviour:

From the root directory of iris:

1. git checkout 7fd7960e25cefc48c49a6e518cfd02f42fa860ea
2. pip install --no-deps --editable .
3. python setup.py test
4. python lib/iris/tests/idiff.py --results iris_image_test_output

Environment

• Iris Version: as per the iris commit SHA
• Matplotlib Version: 3.4.1

Additional context

Click to expand this section...

Add additional verbose information in a collapsible section.

See here for further details.

[bjlittle]

@SciTools/peloton An ideal issues for the peloton to chase down :+1:

[rcomer]

The comment on that test suggests that the grid is supposed to be blue and the colours were caused by a matplotlib bug. 😕

https://github.com/SciTools/iris/blob/83309c32a6c9cfbd4603732a4075bae0f6f07d45/lib/iris/tests/test_mapping.py#L241-L244

[rcomer]

Maybe the fact that this test result has changed is telling us we can now remove the extra handling in iris.plot.outline 🤔 https://github.com/SciTools/iris/blob/83309c32a6c9cfbd4603732a4075bae0f6f07d45/lib/iris/plot.py#L1276-L1280

[rcomer]

History: #109