In batch cell verification, check if there are zero cells

[jtraglia]

Given zero cells, this will throw an exception. We can check for this & return true.

Traceback (most recent call last):
  File "/Users/jtraglia/Projects/jtraglia/consensus-specs/tests/generators/kzg_7594/main.py", line 812, in <module>
    gen_runner.run_generator("kzg_7594", [
  File "/Users/jtraglia/.asdf/installs/python/3.10.4/lib/python3.10/site-packages/eth2spec/gen_helpers/gen_base/gen_runner.py", line 238, in run_generator
    for test_case in tprov.make_cases():
  File "/Users/jtraglia/Projects/jtraglia/consensus-specs/tests/generators/kzg_7594/main.py", line 795, in cases_fn
    for data in test_case_fn():
  File "/Users/jtraglia/Projects/jtraglia/consensus-specs/tests/generators/kzg_7594/main.py", line 241, in case_verify_cell_kzg_proof_batch
    assert spec.verify_cell_kzg_proof_batch(row_commitments, row_indices, column_indices, cells, proofs)
  File "/Users/jtraglia/.asdf/installs/python/3.10.4/lib/python3.10/site-packages/eth2spec/eip7594/mainnet.py", line 5405, in verify_cell_kzg_proof_batch
    return verify_cell_kzg_proof_batch_impl(row_commitments, row_indices, column_indices, cosets_evals, proofs)
  File "/Users/jtraglia/.asdf/installs/python/3.10.4/lib/python3.10/site-packages/eth2spec/eip7594/mainnet.py", line 5269, in verify_cell_kzg_proof_batch_impl
    rli = bls.bytes48_to_G1(g1_lincomb(KZG_SETUP_G1_MONOMIAL[:n], sum_interp_polys_coeff))
  File "/Users/jtraglia/.asdf/installs/python/3.10.4/lib/python3.10/site-packages/eth2spec/eip7594/mainnet.py", line 4395, in g1_lincomb
    assert len(points) == len(scalars)

[kevaundray]

Wouldn't it be better to make the verification algorithm "complete" ? -- ie not have this edge case

[jtraglia]

Wouldn't it be better to make the verification algorithm "complete" ? -- ie not have this edge case

Hmm yes good idea! I've pushed a commit which does this + a new test.

[asn-d6]

Code LGTM. Should this also come with a test vector? Or the test vector was there already?

[jtraglia]

Code LGTM. Should this also come with a test vector? Or the test vector was there already?

Yes, there's already a test vector for this. That's how I noticed the problem.

consensus-specs/tests/generators/kzg_7594/main.py

Lines 239 to 252 in 9be0529

	# Valid: zero cells
	cells, row_commitments, row_indices, column_indices, proofs = [], [], [], [], []
	assert spec.verify_cell_kzg_proof_batch(row_commitments, row_indices, column_indices, cells, proofs)
	identifier = make_id(row_commitments, row_indices, column_indices, cells, proofs)
	yield f'verify_cell_kzg_proof_batch_case_valid_zero_cells_{identifier}', {
	'input': {
	'row_commitments': encode_hex_list(row_commitments),
	'row_indices': row_indices,
	'column_indices': column_indices,
	'cells': encode_hex_list(cells),
	'proofs': encode_hex_list(proofs),
	},
	'output': True
	}