Test `plonk::recursive_verifier::tests::test_size_optimized_recursion` fails about plonky2 HOT 5 CLOSED

You were right, the config.rate_bits should be common_data.config.rate_bits. I fixed that in #386.
Accidentally this would have also caused a subtle bug (also fixed in the PR) if config.zero_knowledge != common_data.config.zero_knowledge, so thanks for finding this!

from plonky2.

Hmm interesting, it's passing for me. Did it happen on a recent copy of main?

That test uses a few CircuitConfigs basically to demonstrate how a speed-optimized proof could be "shrunk" by wrapping it in a proof with different settings, optimized for size. The most important difference is the codeword rate -- our size-optimized proofs use very small rates, so we can run FRI with fewer rounds, at the cost of having to Merklize a huge LDE.

I wonder if we mixed up the two configs somewhere (that of the proof being recursively verified, vs that of the circuit currently being built). I could see that happening, though I'm not sure how to explain the difference on your machine vs mine.

from plonky2.

I think the test passes in release mode but not in debug mode due to the debug_assert_eq.
This debug_assert_eq is wrong I think, I'll try to fix it.

from plonky2.

Oh right! That's good that you caught it; we normally run that test in release mode (it's one of the most compute heavy) so we didn't notice the issue.

from plonky2.

Fixed in #386.

from plonky2.