Bump to GMT 6.2.0rc2

[weiji14]

The second release candidate for GMT 6.2.0 has been released (https://github.com/GenericMappingTools/gmt/releases/tag/6.2.0rc2) on 25 May 2021 and is installable from the conda-forge devel channel using the following command:

conda install -c conda-forge/label/dev gmt=6.2.0rc2

TODO:

• Bump the GMT version to 6.2.0rc2 used in the CI tests (Bump GMT version from 6.2.0rc1 to 6.2.0rc2 in CI #1290)
• Fix failing tests, mostly those documented in How to resolve flaky tests resulting from using a single GMT session #1242 (comment)
  • test_basemap.py (1) (Update fig.basemap, fig.subplot and fig.wiggle baseline images for GMT 6.2.0rc2 #1291)
  • test_subplot.py (4) (Update fig.basemap, fig.subplot and fig.wiggle baseline images for GMT 6.2.0rc2 #1291)
  • test_text.py (18) (Update fig.text baseline images for GMT 6.2.0rc2 #1292)
  • test_wiggle.py (1) (Update fig.basemap, fig.subplot and fig.wiggle baseline images for GMT 6.2.0rc2 #1291)
• Remove xfail pytest markers on tests that are now xpass
  • test_grdtrack.py? (tests still fail on macOS/Windows so don't remove the xfail, see grdtrack produces incorrect results for xarray.DataArray grid #1309)
  • test_grdimage.py (Remove xfails on grdimage test for GMT 6.2.0rc2 #1296)
  • test_makecpt.py ()
  • test_solar.py (Remove xfails on solar and velo tests for GMT 6.2.0rc2 #1297, Remove more xfail from solar tests for GMT 6.2.0rc2 #1305)
  • test_velo.py (Remove xfails on solar and velo tests for GMT 6.2.0rc2 #1297)
• Add some new tests for upstream bugs that were fixed in 6.2.0rc2 (if any)
  • grdimage does not plot subsets of data correctly on a global map #732?

Supersedes #1217

Are you willing to help implement and maintain this feature? Yes

[weiji14]

• Fix failing tests, mostly those documented in #1242 (comment)

  • test_basemap.py (1)
  • test_subplot.py (4)
  • test_text.py (18)
  • test_wiggle.py (1)

Instead of opening 4 separate PRs for each of these, how about making just 2? I.e. one for basemap/subplot/wiggle, and one for text? Do people have any preference?

[seisman]

• Fix failing tests, mostly those documented in #1242 (comment)

  • test_basemap.py (1)
  • test_subplot.py (4)
  • test_text.py (18)
  • test_wiggle.py (1)

Instead of opening 4 separate PRs for each of these, how about making just 2? I.e. one for basemap/subplot/wiggle, and one for text? Do people have any preference?

Sounds good to me.

[seisman]

Here is a minimal example to reproduce the flaky tests in test_makecpt.py. In this example, three figures are generated. The 1st and the 3rd figure uses the same code, so we expect makecpt1.eps and makecpt2.eps are the same, but actually these two images are slightly different. However, if we remove the code for the 2nd figure, makecpt1.eps and makecpt2.eps are the same again.

I think it may mean that the settings for each "figure" are not fully isolated and it looks an upstream GMT issue.

import pygmt

# 1st image
fig = pygmt.Figure()
pygmt.makecpt(cmap="categorical", categorical=True)
fig.colorbar(cmap=True, region=[10, 70, -5, 10], frame=True, position="JBC", V=True)
fig.savefig("makecpt1.eps")

# 2nd image
fig = pygmt.Figure()
fig.coast(region=[0, 10, -20, -10], projection="M6i",  frame=True, land="black")
fig.savefig("other.eps")

# 3rd image (same as the 1st image)
fig = pygmt.Figure()
pygmt.makecpt(cmap="categorical", categorical=True)
fig.colorbar(cmap=True, region=[10, 70, -5, 10], frame=True, position="JBC", V=True)
fig.savefig("makecpt2.eps")

@PaulWessel @meghanrjones Please see if the above example can help you debug GMT 6.2.0rc2.

[seisman]

A bash version to reproduce the issue:

gmt begin

gmt figure makecpt1 pdf
gmt makecpt -Ccategorical -W
gmt colorbar -C -R10/70/-5/10 -B -DJBC

gmt figure other pdf
gmt coast -R0/10/-20/-10 -JM6i -B -Gblack

gmt figure makecpt2 pdf
gmt makecpt -Ccategorical -W
gmt colorbar -C -R10/70/-5/10 -B -DJBC

gmt end

EDIT: I opened an upstream issue GenericMappingTools/gmt#5279 to better track it.