Weird that crew workers are failing on ubuntu with devel R only. I can try to spin up a docker container and test this out interactively (e.g. set up the crew controller to save out logs)

I couldn't reproduce the error on r-devel with rocker/geospatial:devel. However, I got a different test error from that container:

Error (test-tar-terra.R:62:5): tar_terra_vect() works
<tar_condition_run/tar_condition_targets/rlang_error/error/condition>
Error: Error running targets::tar_make()
Error messages: targets::tar_meta(fields = error, complete_only = TRUE)
Debugging guide: https://books.ropensci.org/targets/debugging.html
How to ask for help: https://books.ropensci.org/targets/help.html
Last error message:
    the write() function in tar_format() must not create a directory. Found directories inside the data store where there should only be files: _targets/objects/test_terra_vect_shz

I wonder if this has something to do with differences in GDAL versions? terra::gdal() reports "3.0.4" on this container.

That is a bit strange re the crew workers! Thanks for checking that out in a docker container.

3.0.4 is 4 years old, it makes me think that maybe we should set a minimum version GDAL version in the DESCRPTION under SystemReqs and then also maybe manage that in GH actions, where we can provide a range of GDAL versions

I wonder if this has something to do with differences in GDAL versions? terra::gdal() reports "3.0.4" on this container.

Can confirm writing SOZip shapefile requires GDAL >= 3.7. Not sure if we should have a fallback to using utils::zip() (non-SOZip) in this case, or whether it should be an error/test should be skipped on old GDAL

Still looking into the error with the crew containers. I'm going to see if I can set up the controller to save worker logs in a way that'll be accessible as an artifact from GitHub actions—that should be useful going forward.

GDAL 3.7 was released only about a year ago, so it seems maybe not ideal to require >= 3.7, so it might be a good idea to fallback to utils::zip() with older versions. I can open an issue to discuss this.

Hmm, I'm stumped. Nothing should have changed with those crew tests in this PR! It's hard to debug because targets::tar_test() does too good of a job isolating things and I can't figure out how to print the crew worker logs on GitHub actions.

Well, looks like whatever was going on was a temporary GitHub issue.