stack overflow when running overflow_cache_test_state_cache in debug

[jimmygchen]

Description

Getting this error when running cargo test from the root directory:

thread 'tokio-runtime-worker' has overflowed its stack

Reproduced with cargo nextest as well. It doesn't occur on CI, but on our laptops. i think it may be due to different platform max stack size defaults.

The failing test identified is overflow_cache_test_state_cache

Based on this comment, we might be able to get rid of this test once the state cache is removed in da checker:

// THIS TEST CAN BE DELETED ONCE TREE STATES IS MERGED AND WE RIP OUT THE STATE CACHE

The root cause of the stack overflow seems to be in the blob conversion here:

lighthouse/beacon_node/beacon_chain/src/kzg_utils.rs

Line 32 in 969d12d

.map(|blob| ssz_blob_to_crypto_blob::<E>(blob))

Not sure if there's an easy way around it given that the KZG library accepts Blob containing an owned array. It would be nice if it accepts refs like the peerdas KZG lib, so it doesn't require making a copy of the bytes (which will likely be unfeasible when we increase the blob count)

Version

5.2.1+ (current unstable branch)

[michaelsproul]

Noting that we have another report from a Mac user:

• Beacon block streamer test fails on macOS in debug mode #6647

This does seem to be Mac specific, as we do run this test in debug mode on CI:

lighthouse/.github/workflows/test-suite.yml

Lines 241 to 242 in c042dc1

	- name: Run tests in debug
	run: make nextest-debug

(I forgot we still run things in debug 😅 )