Initial Substrate Header Chain Implementation (#296)
* Add pallet template from Substrate Dev Hub * Clean up un-needed stuff from template * Sketch out dispatchable interface * Introduce notion of finality chain * Add dependencies which were removed during a rebase * Sketch out idea for finality header-chain pallet * Sketch out ChainVerifier trait * Add storage parameter to verifier * Write out some things I think I need for finality verification * Add some pseudocode for marking finalized headers * Remove parity_scale_codec duplicate * Move verification logic into pallet I've been struggling with getting the generic types between the storage and verifier traits to play nice with each other. As a way to continue making progress I'm moving everything to the pallet. This way I hope to make progress towards a functional pallet. * Start doing verification around authority set changes * Remove commented BridgeStorage and ChainVerifier traits * Create Substrate bridge primitives crate * Add logic for updating scheduled authority sets * Introduce notion of imported headers * Implement basic header ancestry checker * Add mock runtime for tests * Add testing boilerplate * Add some storage read/write sanity tests * Add some basic header import tests * Add tests for ancestry proofs * Create helper for changing authority sets * Fix authority set test Fixes a problem with how the scheduled change was counted as well as a SCALE encoding issue * Correctly check for scheduled change digests There's no guarantee that the consensus digest item will be the last one in a header, which is how it was previously being checked. Thanks to Andre for pointing me to the Grandpa code that does this. * Mark imported headers as finalized when appropriate When a header that finalizes a chain of headers is succesfully imported we also want to mark its ancestors as finalized. * Add helper for writing test headers * Add test helper for scheduling authority set changes * Bump Substrate pallet and primitives to rc6 * Remove Millau verifier implementation * Add some doc comments * Remove some needless returns * Make Clippy happy * Split block import from finalization * Make tests compile again * Add test for finalizing header after importing children * Create a test stub for importing future justifications * Start adding genesis config * Reject justifications from future We should only be accepting justifications for the header which enacted the current authority set. Any ancestors of that header which require a justification can be imported but they must not be finalized. * Add explanation to some `expect()` calls * Start adding GenesisConfig * Plug genesis config into runtime * Remove tests module * Check for overflow when updating authority sets * Make verifier take ownership of headers during import * Only store best finalized header hash Removed the need to store the whole header, since we store it was part of the ImportedHeaders structure anyways * Add some helpers to ImportedHeader * Update ancestry checker to work with ImportedHeaders * Update ancestry tests to use ImportedHeaders * Update import tests to use ImportedHeaders * Clean up some of the test helpers * Remove stray dbg! * Add doc comments throughout * Remove runtime related code * Fix Clippy warnings * Remove trait bound on ImportedHeader struct * Simplify checks in GenesisConfig * Rename `get_header_by_hash()` * Alias `parity_scale_codec` to `codec` * Reword Verifier documentation * Missed codec rename in tests * Split ImportError into FinalizationError * Remove ChainVerifier trait This trait was a remenant of the original design, and it is not required at the moment. Something like it should be added back in the future to ensure that other chains which conform to this interface can be used by higher-level bridge applications. * Fix the verifier tests so they compile * Implement Deref for ImportedHeader * Get rid of `new` methods for some Substrate primitives * Ensure that a child header's number follows its parent's * Prevent ancestry checker from aimlessly traversing to genesis If an ancestor which was newer than the child header we were checking we would walk all the way to genesis before realizing that we weren't related. This commit fixes that. * Remove redundant clones * Ensure that old headers are not finalized Prevents a panic where if the header being imported and `best_finalized` were the same header the ancestry checker would return an empty list. We had made an assumption that the list would always be populated, and if this didn't hold we would end up panicking. * Disallow imports at same height as `best_finalized` * Fix Clippy warnings * Make NextScheduledChange optional * Rework how scheduled authority set changes are enacted We now require a justification for headers which _enact_ changes instead of those which _schedule_ changes. A few changes had to be made to accomodate this, such as changing when we check for scheduled change logs in incoming headers. * Update documentation for Substrate Primitives * Clarify why we skip header in requires_justification check * Add description to assert! call * Fix formatting within macros * Remove unused dependencies from runtime * Remove expect call in GenesisConfig * Turn FinalityProof into a struct * Add some inline TODOs for follow up PRs * Remove test which enacted multiple changes This should be added back at some later point in time, but right now the code doesn't allow for this behaviour. * Use `contains_key` when checking for header This is better than using `get().is_some()` since we skip decoding the storage value * Use initial hash when updating best_finalized * Add better checks around enacting scheduled changes * Rename finality related functions * Appease Clippy
parent
b8fa003c
Please register or sign in to comment