Rust API test_arrow test fails due to Utf8Error on macos-11 GitHub-hosted runner

[russell-liu]

The test has failed 2 out of 2 times so far.
Logs and workflow files can be found here:
https://github.com/kuzudb/kuzu/actions/runs/6111754214/job/16587694174
https://github.com/kuzudb/kuzu/actions/runs/6113068279/job/16591808577

[russell-liu]

I think I've figured out why it is failing -- the macos-11 runner uses Clang/LLVM 13.0.0 by default (https://github.com/actions/runner-images/blob/main/images/macos/macos-11-Readme.md), whereas we inexplicitly require 14. This is demonstrated by these two runs, one of which failed and one of which succeeded, which differ only by Clang/LLVM version:
https://github.com/kuzudb/kuzu/actions/runs/6122821358/job/16619417254
https://github.com/kuzudb/kuzu/actions/runs/6122418640/job/16618126207

[russell-liu]

Never mind, it can also fail with Clang/LLVM 14:
https://github.com/kuzudb/kuzu/actions/runs/6122975487/job/16619878211

[benjaminwinger]

As some context for the error, it's being raised by this line in arrow-rs, where it's accessing the name field of the schema.

That name field should either be what's set here as a hardcoded ascii string, or one of the property names as set here (the rust test failing has a string column called name and an int64 column called age).

I think the problem is that the property names are owned by the DataTypeInfo struct, rather than being copied into the ArrowSchema like the rest of the information, and the DataTypeInfo used in getArrowSchema is temporary. Struct field names are copied into the ArrowSchemaHolder, but property names are not.

We should also add a tests of getArrowSchema and getNextArrowChunk, as issues like this could be caught by the address sanitizer (it may not be needed for the rust code, but it is needed for C++ interfaces used by the rust code).

[russell-liu]

It also happens on Windows as well:
https://github.com/kuzudb/kuzu/actions/runs/6189154608/job/16802666644

[benjaminwinger]

Yes, it can occur on all platforms, but I think the last time you updated the branch you're testing on was the commit before #2019 was merged.

[russell-liu]

Oops, I had this page open and forgot to refresh it, so I wasn't aware that it has been fixed. Please ignore my previous comment.