[Karinon]

Tackling the issues #68, #42, #41 and #40

This updates the build dependencies to current releases and adds support for more compressed NetCDF/HDF5 files.

What changed

• Updated versions: Emscripten 6.0.13, HDF5 2.2.0 and NetCDF-C 4.10.1.
• New compression filters: files compressed with Blosc, Zstandard or bzip2 can now be read. WebAssembly can't load HDF5 filter plugins at runtime, so these filters are built into the WASM module and registered when it starts.
• Simpler build script: the newer NetCDF and HDF5 versions build cleanly with Emscripten, so we no longer patch CMakeLists.txt and config.h by hand.
• Rebuilds on version changes: the build script now detects when installed dependencies are from a different version and rebuilds them, so stale caches aren't reused.

Notes

• The CI cache key was bumped, so the first CI run will rebuild everything from scratch.

Tried it out with gridlook as well as the integrated viewer.

[TheJeran]

Do you have any zstandard NCs to test this? The one from EarthyScience/Browzarr#549 expired.

[lazarusA]

I know that this is a bit out of scope here. But, we could we see if this also builds on CI for macOS, or if it's preferred in a follow-up PR. I usually test this locally on macOS, but today I cannot verify this build locally, so, seeing it on CI will make this more mergeable. Windows probably will be trickier or easier, not sure if with this new build there are not more issues. My last attempt was not very successful #34

[Karinon]

Do you have any zstandard NCs to test this? The one from EarthyScience/Browzarr#549 expired.

I don't have any real-world-examples, but I created a synthetic one on a regular grid. Didn't work correctly in the previous release, but does now.

https://swift.dkrz.de/v1/dkrz_a073cd57eae14eba9afc9c692de6c127/Foo/global-zstd.nc?temp_url_sig=44d5911d1c82bf07d613bc4d5b6e35910ff4f111&temp_url_expires=2026-11-29T11:52:49Z

[Karinon]

I know that this is a bit out of scope here. But, we could we see if this also builds on CI for macOS, or if it's preferred in a follow-up PR. I usually test this locally on macOS, but today I cannot verify this build locally, so, seeing it on CI will make this more mergeable. Windows probably will be trickier or easier, not sure if with this new build there are not more issues. My last attempt was not very successful #34

At least MacOS was quite easy to integrate into the CI (checked it on my personal fork beforehand). Windows is another kind of beast and should definitely be a separate PR