Ward Fisher

I'm observing the following when running with `1.12.0`: ```` 73/211 Test #73: nc_test_test_byterange ................***Failed 21.02 sec *** Testing reading NetCDF-3 file with http ***Test remote classic file ***Test remote netdf-4...

Don't trust the Github Actions yet, we aren't testing against 1.12 there (yet), so I would expect them to pass. Can you recreate the issue locally?

Refresh my memory if you would, @DennisHeimbigner, is there a minimum required hdf5 version for byterange support?

Possibly. The failure in github actions is specific to HDF5 1.10.1, so we can ignore that for the moment. I'll try to figure out what's going on, it might also...

Rebuilding HDF5 `1.12.0` with debugging and sanitization enabled to try to make sense of this; also need to test on x64 architecture, the failures thus far have been on ARM.

Ok, it looks like the cmake-based tests aren't running parallel szlib tests, even though (at a glance) they should be. I will dig into this further in the AM.

Got things sorted out with the build system, off to the issue itself. Seeing the following via cmake or autotools with `run_par_test.sh`: ```` *** testing szip compression with parallel I/O...malloc():...

This is a problem, I will see what we can do.

I've added a note, and will add the fix. It's not as disastrous as my initial impression since we still have the symbolic links, so it's also not great. But...

If you run `ls -alh /mathworks/devel/sandbox/abaruah/3p-tmw/3p/derived/glnxa64/netcdf/ncdap_test/tst_remote3.sh`, what do you see? This seems like a pretty straightforward linux filesystem permissions issue, and it doesn't appear to be reflecting anything specific to...