[kriss39]

Summary

• register the PQ verifier precompile only when Zero6 is active
• cover both pre-Zero6 and post-Zero6 availability

Testing

• cargo fmt --check
• cargo test -p arc-precompiles precompile_provider

Fixes #293

[kriss39]

@ZhiyuCircle gentle ping on this one when you have time.

[osr21]

One additional regression would help verify the fork wiring, not just the precompile-map branch: the current tests construct ArcHardforkFlags directly, so they do not show that block execution derives and passes the correct flags from each block's chain-spec activation.

Because precompile registration is selected per block, please consider testing a pre-Zero6 block and the first Zero6-active block using chain-spec-derived activation (or the closest integration-level fixture). Zero6 is active from genesis on mainnet, but activated later on testnet and devnet; historical execution of pre-activation blocks must retain the earlier precompile set. This also guards against a future change that accidentally makes PQ registration unconditional.

[kriss39]

Thanks, good point. Added factory_registers_pq_precompile_from_block_derived_zero6 in ba2a57b: it builds a localdev chain spec with Zero6 at block 10 and creates the EVM through ArcEvmFactory (both create_evm and create_evm_with_inspector), so the flags come from get_hardfork_flags on the block env. PQ is absent at block 9 and present at block 10. With the gate reverted the test fails at block 9.

[osr21]

Thanks, this closes the key wiring gap: the test derives Zero6 from the block environment and checks both factory construction paths on either side of the activation boundary. One optional follow-up, if production chain-spec fixtures are readily available, would be to assert the configured boundary for each network (including mainnet’s genesis activation and pre-activation historical blocks on networks where it activates later). That would protect the network schedules as well as the factory wiring; if those schedules are already covered elsewhere, the localdev block-9/block-10 regression looks well scoped for this PR.