Review FAILED tests

[ES-Solar]

When I run tests on the master branch in macos, it is reporting 91 failed tests. For the first couple I looked at, the resulting ps files look good or maybe better (better formatting) than the original baseline. It would be helpful to better understand the failures and determine which actually represent problems.
So I'm creating this issue to help document my review of the chronically failing tests.

[github-actions]

👋 Thanks for opening your first issue here! Please make sure you filled out the template with as much detail as possible. We appreciate that you took the time to contribute!

Please make sure you read our Contributing Guide and abide by our Code of Conduct.

[seisman]

Have you set GMT_DATA_SERVER to static when you run the tests locally? The CI uses the static server, and there are only a few failing tests.

[joa-quim]

Extending a bit more what Dongdong said, we have the DVC exactly because of this. Before we used to have the PS files directly included in the GMT repo, but that was making the repo's size increase fast and not stop. That happened, for example, whenever data from one of the hosted data sources (earth_relief, for example) was updated. Because of this we have a collection of remote files used in tests that are not updated. To use them, the GMT_DATA_SERVER must be set to static.

[ES-Solar]

I was not aware of needing to set the GMT_DATA_SERVER variable to static, thanks for the guidance.

[ES-Solar]

I reran the test with the GMT_DATA_SERVER set to static and got the same results, 91 failures. Is that not close to what youwould expect?
I am including the examples in the docs as well as the 'tests', ie:
set (DO_EXAMPLES TRUE)
set (DO_TESTS TRUE)

[joa-quim]

To be honest, I also have troubles with that GMT_DATA_SERVER setting and since I've no patience to fight scripts, I set
set (GMT_DATA_SERVER static)
in ConfigUser.cmake before running the tests.

[MarceloBanik]

changing line 212 in ConfigUserAdvanced.cmake from

# set (GMT_DATA_SERVER "static")

to

set (GMT_DATA_SERVER "static")

I went from 80 tests failed out of 1194 to 31 tests failed out of 1194.

[ES-Solar]

Thanks for the comment, I had changed that line but I must have missed something in getting the change to be picked up in the ctest run. I'll review my re-building procedures

[ES-Solar]

I think I've got the server settings and I enabled all examples and supplemental test to match what I saw in the CI configuration so now I see 41 failures out of 1300 tests. Several of the first few tests seem to fail because of differences in how the resulting images are being cropped by psconvert.
I've found this line in the gmt_modern.h file:
#define GMT_SESSION_CONVERT "A" /* Default psconvert options in gmt figure */
I'm wondering if this option was in some update after the original baseline images were generated. Perhaps I should just update the baseline images with the auto-cropped versions?

[seisman]

what's your gs version?

[joa-quim]

Maybe the gs, maybe the baseline images are not what they should be. We (I?) and de CI have not those failures so the solution is not to change the baseline PSs.

[ES-Solar]

I have GPL Ghostscript 10.05.1 (2025-04-29)

[Esteban82]

I have GPL Ghostscript 10.05.1 (2025-04-29)

I had some issues with gs 10.05 (#8840). Try using 10.06 instead.

[ES-Solar]

Also, I thought that the CI tests for the PRs typically fail because of problems with the dvc pull request. For example, the PR for "Bump actions/checkout from 6.0.1 to 6.0.2 #2894" reports a failure for "Pull baseline image data from dvc remote" for all 3 platforms. So, I'm confused about where the CI tests are passing. I think I'm missing something fundamental in how the automated workflow is working and how to investigate the errors.