[iris.ci] environment lockfiles auto-update

[scitools-ci]

Lockfiles updated to the latest resolvable environment.

If the CI tasks fail, create a new branch based on this PR and add the required fixes to that branch.

[bjlittle]

⚡ wow!

[bjlittle]

Oof!

2022-09-15T13:00:46.0903981Z make[1]: *** [Makefile:45: html] Segmentation fault (core dumped)

[trexfeathers]

I am at least happy with the changes themselves - no unexpected packages etc.

[trexfeathers]

The failure is rooted in: SciTools/cartopy@8860a81 - this change has been planned for several years, we just never spotted the warning (see #4384). It was finally implemented in SciTools/cartopy@fcb784d.

The change makes the below expected_mask incorrect. It was created as a 'snapshot' rather than an explicitly desired outcome, so should be fine to change it; will take a look now.

iris/lib/iris/tests/unit/analysis/cartography/test_rotate_winds.py

Lines 422 to 435 in 06c8bc6

	# Snapshot of mask with fixed tolerance of atol=2e-3
	expected_mask = np.array(
	[
	[1, 1, 1, 0, 0, 0, 0, 0, 0, 1],
	[1, 1, 1, 0, 0, 0, 0, 0, 0, 1],
	[1, 1, 1, 1, 0, 0, 0, 0, 1, 1],
	[1, 1, 1, 1, 0, 0, 0, 0, 1, 1],
	[1, 1, 1, 1, 0, 0, 0, 0, 1, 1],
	[1, 1, 1, 1, 1, 0, 0, 1, 1, 1],
	[1, 1, 1, 1, 1, 0, 0, 1, 1, 1],
	[1, 1, 1, 1, 1, 0, 0, 1, 1, 1],
	],
	np.bool_,
	)

[trexfeathers]

It was created as a 'snapshot' rather than an explicitly desired outcome, so should be fine to change it; will take a look now.

I'm no longer confident to make a change. This is all about testing whether masking removes deltas above a certain threshold, and with the underlying Cartopy changes this is no longer true. Drastically reducing the tolerance from 2e-3 to 2e-5 allows the below test to pass, but if the test sample size is increased it fails again.

iris/lib/iris/tests/unit/analysis/cartography/test_rotate_winds.py

Line 446 in 06c8bc6

self.assertTrue(anom[~ut.data.mask].max() < 0.1)

The below statement sounds like the original tolerance was a shot in the dark somewhat, and reading the conversations in the original PR - #1546 - backs this up.

iris/lib/iris/analysis/cartography.py

Lines 1011 to 1012 in 06c8bc6

	# A value of atol=2e-3 is used, which corresponds to a change in magnitude
	# of approximately 0.1%.

What is the failing test achieving? Just monitoring that the behaviour remains the same? If so what should the expected results be changed to? Should the masking tolerance also be changed?

[trexfeathers]

Note that the following change allows all tests to pass, but is basically pinning us to old behaviour instead of adopting Proj's latest.

diff --git a/lib/iris/coord_systems.py b/lib/iris/coord_systems.py
index 802571925..ce00746ee 100644
--- a/lib/iris/coord_systems.py
+++ b/lib/iris/coord_systems.py
@@ -736,10 +736,10 @@ class OSGB(TransverseMercator):
         )
 
     def as_cartopy_crs(self):
-        return ccrs.OSGB()
+        return ccrs.OSGB(True)
 
     def as_cartopy_projection(self):
-        return ccrs.OSGB()
+        return ccrs.OSGB(True)

[trexfeathers]

Superceded by #4968.