[C++] Add new datatypes to the IPC fuzz seed corpus

[pitrou]

Describe the enhancement requested

Part of our seed corpus for IPC fuzzing is generated dynamically here, but it hasn't been updated for the newer datatypes and features:

arrow/cpp/src/arrow/ipc/generate_fuzz_corpus.cc

Line 68 in f7530de

Result<std::vector<std::shared_ptr<RecordBatch>>> Batches() {

We should add the newer datatypes (run-end encoding, list view, string view) and ideally also new IPC features such as optional buffer compression (perhaps generate both compressed and uncompressed files).

Component(s)

C++, Continuous Integration

[pitrou]

@felipecrv This may be interesting to you.

[pitrou]

Note that you can check your work by running this nightly build, or reproducing its steps locally:
https://github.com/ursacomputing/crossbow/actions/runs/6413041512/job/17411340300

For example you can see the dynamic generation occurring here:
https://github.com/ursacomputing/crossbow/actions/runs/6413041512/job/17411340300#step:6:832

[pitrou]

And there are docs here: https://arrow.apache.org/docs/dev/developers/cpp/fuzzing.html :-)

[pitrou]

@zeroshade Can we prioritize this for 0.15.0? This a rather small task and is important to ensure robustness of the newly added data type implementations.

[felipecrv]

I would like to work on this.

[pitrou]

@felipecrv Are you still interested in this? This is really something that we should do, otherwise we may be missing out on potential reliability issues.

[felipecrv]

I'm assigning myself. I will at least get this started.

[pitrou]

Ping @felipecrv :-)

[felipecrv]

Ping @felipecrv :-)

With all the reviews and Flight/SQL/REST work, I don't have time to take another workstream :/