use stable rust

[alexander-buerger-met-no]

• use hidefix 0.11.1
• use divan for benches

[alexander-buerger-met-no]

@magnusuMET I am not sure how to avoid the make_extents function in dars/src/lib.rs. This function is copied from the hidefix tests for Extents.

[magnusuMET]

I don't see an easy way since type inference through the result type is tricky. You might make stream_xdr take an Extents instead, pushing the trouble of making Extents upwards in the stack

[magnusuMET]

The memory leak is from H5MM_xstrdup, found by running

env RUSTFLAGS="-Z sanitizer=address" ASAN_SYMBOLIZER_PATH=/usr/bin/llvm-symbolizer-14 cargo test --target x86_64-unknown-linux-gnu -p dars --test coads -- --nocapture coads_subset

[gauteh]

The memory leak is from H5MM_xstrdup, found by running

env RUSTFLAGS="-Z sanitizer=address" ASAN_SYMBOLIZER_PATH=/usr/bin/llvm-symbolizer-14 cargo test --target x86_64-unknown-linux-gnu -p dars --test coads -- --nocapture coads_subset

Interesting, probably of interest to hdf5 as well.

[magnusuMET]

We run the address sanitizer in the hdf5 crate which does not reveal any leaks. The backtrace isn't very helpful either and I am not sure how to track down the leak

[gauteh]

Maybe it is caused by dars, some of the attribute stuff is using hdf directly.

ons. 14. aug. 2024, 11:01 skrev Magnus Ulimoen @.***>:

We run the address sanitizer in the hdf5 crate which does not reveal any leaks. The backtrace isn't very helpful either and I am not sure how to track down the leak

— Reply to this email directly, view it on GitHub https://github.com/gauteh/dars/pull/37#issuecomment-2288228332, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAN3654MHJF6CUFWNKWKVLZRMMFPAVCNFSM6AAAAABMONX3IOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEOBYGIZDQMZTGI . You are receiving this because you commented.Message ID: @.***>

[magnusuMET]

Reading data/dmrpp/unused/ntypes.h5 gives the memory leak, in specific the compound_type_4 dataset

[gauteh]

Reading data/dmrpp/unused/ntypes.h5 gives the memory leak, in specific the compound_type_4 dataset

Those files were found in the hyrax test suite. They also appear to be less unused than the path suggests.

[magnusuMET]

I removed that troublesome file. If someone wants to hunt down that leak we can always add it back in.

Clippy was also really strict for package metadata, if we are planning on releasing the crates on crates.io we could blacklist the warnings again.

[gauteh]

Great! Thanks a lot both!