eip7549: Ensure non-zero bits for each committee bitfield comprising an aggregate

[mkalinin]

Currently, the spec allows an on chain aggregate to have only a single non-zero bit regardless of the number of committees. This allows to create an on chain aggregate with 63 committees worth of zeroes which successfully passes all validations (~4kb of uncompressed and wasteful data per aggregate).

This PR adds a validation that requires at least a single non-zero bit for each attesting committee bitfield comprising the aggregate.

[jtraglia]

Good idea, I agree we should check for this. LGTM, with one little nit.

[jtraglia]

Do we need to add another condition to here?

consensus-specs/specs/electra/p2p-interface.md

Lines 51 to 53 in 1b408e9

	The following validations are added:
	* [REJECT] `len(committee_indices) == 1`, where `committee_indices = get_committee_indices(aggregate)`.
	* [REJECT] `aggregate.data.index == 0`

[arnetheduck]

This allows to create an on chain aggregate with 63 committees worth of zeroes which successfully passes all validations (~4kb of uncompressed and wasteful data per aggregate).

what if a whole committee is not voting? ie there are no attestations from that commiteee?

[ensi321]

This allows to create an on chain aggregate with 63 committees worth of zeroes which successfully passes all validations (~4kb of uncompressed and wasteful data per aggregate).

what if a whole committee is not voting? ie there are no attestations from that commiteee?

The validator who does aggregation should include its own vote at the very least. So there should be at least one bit in aggregation bits.

edit: I notice https://github.com/ethereum/consensus-specs/blob/dev/specs/phase0/validator.md#construct-aggregate says aggregator should collect all the votes (including its own) that come solely from gossip. But I believe aggregator should include its own vote even if it doesn't receive its own vote through gossip.

[mkalinin]

This allows to create an on chain aggregate with 63 committees worth of zeroes which successfully passes all validations (~4kb of uncompressed and wasteful data per aggregate).

what if a whole committee is not voting? ie there are no attestations from that commiteee?

Then the bit corresponding to this committee should be 0 in the committee_bits. The case this PR prevents is when a committee bit is 1 while aggregation_bits corresponding to this committee are all zeroes.

[mkalinin]

Do we need to add another condition to here?

consensus-specs/specs/electra/p2p-interface.md

Lines 51 to 53 in 1b408e9

	The following validations are added:
	* [REJECT] `len(committee_indices) == 1`, where `committee_indices = get_committee_indices(aggregate)`.
	* [REJECT] `aggregate.data.index == 0`

Note that a network aggregate only allows a single committee, so the following Phase0 check is enough:

consensus-specs/specs/phase0/p2p-interface.md

Lines 371 to 372 in 49b6840

	- _[REJECT]_ The aggregate attestation has participants --
	that is, `len(get_attesting_indices(state, aggregate)) >= 1`.