We have a set of test failures that are long-lived or due to sutble architectural differences between Windows and *nix. These are marked with # GMT_KNOWN_FAILURE_WINDOWS or # GMT_KNOWN_FAILURE. Since we now never get reminders that these fail, we forget.

One can tell Cmake to include these failures by uncommenting the line

set (GMT_ENABLE_KNOWN2FAIL OFF)

in cmake/ConfigUserAdvanced.cmake and build and run the tests. I just did that. That way I relearned that, for instance, psbasemap/map_JE.sh is flagged as a Windows failure but the plot looks fine on macOS so I assume these are tiny differences in the gridlines that triggers the failures. That is very different from actual failures across all OS due to bugs in GMT. Just looking at the PNG and PDFs resulting from the failed tests I wonder if these non-issues but failing Windows test scripts:

psbasemap/map_JE.sh
test/grdvector/vectors.sh
test/mapproject/oblmerc_down.sh

should instead be reclassified with a slightly higher RMS setting, e.g. as in

./pscoast/oblsuite.sh:# GRAPHICSMAGICK_RMS = 0.02

@meghanrjones, could you please have a look at this and determine what those RMS might be and then we can remove the GMT_KNOWN_FAILURE_WINDOWS and instead set the higher RMS trigger points?