reject invalid proofs when validating blobs

[etan-status]

Currently, passing 0xc00000... proof seems to pass verifyProofs. Unsure why such a check is not necessary in spec, and also unsure whether it is correct to reject proof at infinity, or if it could occur, e.g., for a blob containing all 0 bytes. Weird overall...

[hwwhww]

It is actually a valid case in https://github.com/ethereum/consensus-spec-tests/blob/master/tests/general/deneb/kzg/verify_blob_kzg_proof/kzg-mainnet/verify_blob_kzg_proof_case_correct_proof_0951cfd9ab47a8d3/data.yaml

[asn-d6]

Proof being the point at infinity is a valid case for when the blob is all zeroes. A blob being all zeroes is basically the zero polynomial -- its commitment will be the point at infinity and the corresponding proof will also be the point at infinity.

A related edge case can happen for a blob full of twos (or any other constant number): A blob full of twos would correspond to the polynomial p(x) = 2 and it's corresponding KZG proof would be the point at infinity. It's corresponding commitment would be 2G (not the point at infinity).

If these edge cases are spilling out to clients and causing confusion, we should probably document them somewhere.

[github-actions]

Unit Test Results

         9 files  ±0    1 089 suites  ±0   45m 12s ⏱️ + 5m 15s
  3 847 tests ±0    3 568 ✔️ ±0  279 💤 ±0  0 ❌ ±0 
15 892 runs  ±0  15 587 ✔️ ±0  305 💤 ±0  0 ❌ ±0 

Results for commit 905ad2e. ± Comparison against base commit ee75d45.

[etan-status]

Found the problem. There are two output values, one is a return value, and the other is a bool *. We only checked the first one but not the second. When I check both, the test works fine :)

[etan-status]

Introduction of problem here: #4732

Was always broken. Nice that the tests caught it 😅