Fix test generator duplicate key issue

[hwwhww]

Fix #2650

Issue

We have key conflicts in the genesis and sanity test generators.

• consensus-specs/tests/generators/genesis/main.py

  Lines 6 to 15 in fea3702

  	phase_0_mods = {key: 'eth2spec.test.phase0.genesis.test_' + key for key in [
  	'initialization',
  	'validity',
  	]}
  	altair_mods = phase_0_mods
  	# we have new unconditional lines in `initialize_beacon_state_from_eth1` and we want to test it
  	merge_mods = {
  	**{key: 'eth2spec.test.merge.genesis.test_' + key for key in [
  	'initialization',
  	]},

• consensus-specs/tests/generators/sanity/main.py

  Lines 10 to 15 in fea3702

  	altair_mods = {**{key: 'eth2spec.test.altair.sanity.test_' + key for key in [
  	'blocks',
  	]}, **phase_0_mods}
  	merge_mods = {**{key: 'eth2spec.test.merge.sanity.test_' + key for key in [
  	'blocks',
  	]}, **altair_mods}

In these test generators, the handler names were conflicting. Thus it did NOT generate the test vectors of the newer file.

How did I fix it

Add combine_mods helper to deal with the test sources.

• p.s. usually I'd name it merge_mods, but you know, too confusing.

TODO

• I got errors when I generated sanity tests:

➜ make detect_generator_incomplete
find ../consensus-spec-tests/tests -name "INCOMPLETE"
../consensus-spec-tests/tests/mainnet/merge/sanity/blocks/pyspec_tests/half_sync_committee_committee/INCOMPLETE
../consensus-spec-tests/tests/mainnet/merge/sanity/blocks/pyspec_tests/half_sync_committee_committee_genesis/INCOMPLETE
../consensus-spec-tests/tests/mainnet/altair/sanity/blocks/pyspec_tests/half_sync_committee_committee/INCOMPLETE
../consensus-spec-tests/tests/mainnet/altair/sanity/blocks/pyspec_tests/half_sync_committee_committee_genesis/INCOMPLETE

Update: fixed in 4ae8fb1

[hwwhww]

The previous sync_committee_bits would be incorrect if there are duplicate indices in committee.

[hwwhww]

Previously

if committee = [1, 2, 3, 1]
participants = [2, 1] <— that means the the 2nd and 4th items were selected, not 1st+2nd+4th

sync_committee_bits=[index in participants for index in committee] would be return [True, True, False, True]

The aggregate signature from compute_aggregate_sync_committee_signature would only include the 2nd sig once + 4th sig once rather than the duplicate sigs.

This fix

The new fix generates the correct sync_committee_bits first and then calculate the corresponding participants.

[ralexstokes]

ah! great catch! left some feedback

[ralexstokes]

nice work! some minor feedback above