Signed-off-by: Adrian Catangiu <[email protected]>

Signed-off-by: Adrian Catangiu <[email protected]>

Signed-off-by: Adrian Catangiu <[email protected]>

Signed-off-by: Adrian Catangiu <[email protected]>

GRANDPA finalization proof is not checked, which leads to slashing on
forks. This is fine since honest validators will not be slashed on the
chain finalized by GRANDPA, which is the only chain that ultimately
matters. The only material difference not checking GRANDPA proofs makes
is that validators are not slashed for signing BEEFY commitments prior
to the blocks committed to being finalized by GRANDPA. This is fine too,
since the slashing risk of committing to an incorrect block implies
validators will only sign blocks they *know* will be finalized by
GRANDPA.

instead of using votes as the underlying primitive, rather use
commitments since they're a more universal container for signed payloads
(for instance `SignedCommitment` is also the primitive used by ethereum relayers).
SignedCommitments are already aggregates of multiple signatures. Will
use SignedCommitment directly next.

previously assumed equivocation report for singular malicious party.
With fork equivocations, the expectation should be that most
equivocation reports will be for multiple simultaneous equivocators.

reduce from Block to Header: less restrictive.

check_signed commitment wasn't complete anyway.
good to have both interfaces since BeefyVersionedFinalityProof is what's
passed on the gossip layer, and SignedCommitments are naturally
reconstructed from multiple sources, for instance submit_initial calls
on Ethereum.

redundant vs report_fork_equivocation

No need to trigger first session if chain's already had blocks built,
such as with generate_blocks_and_sync, which needs to build on genesis.
If chain is not at genesis and create_beefy_worker, it will panic trying
to canonicalize an invalid (first) block.

can pass in Arc of TestApi now. Required since fisherman reports should be
pushed to `reported_fork_equivocations`, and should be logged with the
same instance (see upcoming commit).

mock api's `submit_report_fork_equivocation_unsigned_extrinsic` pushes
reported equivocations to runtime_api.reported_fork_equivocations

Generates fork equivocation proofs from a vote and a header.

i.e. a fork equivocation triggered by a vote

required for fisherman to *not* report own equivocations - see next commit.

i.e. a fork equivocation triggered by a vote

Co-authored-by: Adrian Catangiu <[email protected]>

skip weight test for now

else fork equivocation via signed commitment tests will be less
meaningful (no validators that don't equivocate, ergo don't get slashed)

Already reported equivocators should neither be slashed again, nor
should they cause the report containing a re-report of their
equivocation to be invalidated by their presence.

…-fisherman

stick with Vec since now in practice, always iterating through entire
signatories & not performing any key lookup.

also leaving ForkEquivocationProof { commitment, signatories, .. } as is
for now - let's change to SignedCommitment if it gets in the way.

…-fisherman

…-fisherman

…-fisherman

…lashing-fisherman

…-fisherman

Co-authored-by: Serban Iorga <[email protected]>

Reported-by: Serban Iorga <[email protected]>

Co-authored-by: Serban Iorga <[email protected]>

Co-authored-by: Adrian Catangiu <[email protected]>