feat: Add support for CommitDiff to efficiently commit small changes in large delegated accounts
#575
Conversation
This stack of pull requests is managed by Graphite. Learn more about stacking. |
WalkthroughAdds a CommitTask constructor that may prefetch a base account and choose commit-diff vs commit-state at runtime; introduces CommittedAccount; integrates a zero-copy OrderBook and new order-book instructions/handlers; threads per-context user_seed through client and tests; updates manifests and many tests; bumps delegation-program rev. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant Caller
participant CommitTask
participant RpcClient
participant DelegationProgram
Caller->>CommitTask: CommitTask::new(commit_id, allow_undelegation, committed_account)
CommitTask->>CommitTask: inspect committed_account.account.data.len()
alt data_len > COMMIT_STATE_SIZE_THRESHOLD
CommitTask->>RpcClient: fetch base account (RPC)
RpcClient-->>CommitTask: Account (fetched)
CommitTask->>CommitTask: store fetched_account
else
CommitTask->>CommitTask: proceed without fetched_account
end
Caller->>CommitTask: create_commit_ix(validator)
alt is_commit_diff()
CommitTask->>CommitTask: create_commit_diff_ix(validator, fetched_account)
CommitTask->>DelegationProgram: invoke commit_diff
else
CommitTask->>CommitTask: create_commit_state_ix(validator)
CommitTask->>DelegationProgram: invoke commit_state
end
CommitTask-->>Caller: Instruction
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Areas to focus on:
Possibly related PRs
Suggested reviewers
Pre-merge checks and finishing touches✅ Passed checks (2 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
snawaz
force-pushed
the
snawaz/commit-diff
branch
6 times, most recently
from
249b1aa to
2f0555f
Compare
snawaz
force-pushed
the
snawaz/commit-diff
branch
2 times, most recently
from
0151da0 to
5c25e73
Compare
snawaz
changed the title
MagicBlockInstruction::ScheduleCommitDiffAndUndelegate to efficiently commit changes in delegated accounts
snawaz
changed the title
MagicBlockInstruction::ScheduleCommitDiffAndUndelegate to efficiently commit changes in delegated accountsScheduleCommitDiffAndUndelegate to efficiently commit changes in delegated accounts
snawaz
force-pushed
the
snawaz/commit-diff
branch
2 times, most recently
from
2ae750b to
725f72c
Compare
Manual Deploy AvailableYou can trigger a manual deploy of this PR branch to testnet: Alternative: Comment
Comment updated automatically when the PR is synchronized. |
snawaz
force-pushed
the
snawaz/commit-diff
branch
from
725f72c to
e5c5e15
Compare
snawaz
changed the title
ScheduleCommitDiffAndUndelegate to efficiently commit changes in delegated accountsScheduleCommitDiffAndUndelegate to efficiently commit changes in delegated accounts
snawaz
changed the title
ScheduleCommitDiffAndUndelegate to efficiently commit changes in delegated accountsScheduleCommitDiffAndUndelegate to efficiently commit changes in delegated accounts
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 29
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
magicblock-committor-service/src/tasks/args_task.rs (1)
229-237: reset_commit_id ignores CommitDiff—update both variants.Commit IDs must update for Commit and CommitDiff.
Apply:
- let ArgsTaskType::Commit(commit_task) = &mut self.task_type else { - log::error!("reset_commit_id"); - return; - }; - - commit_task.commit_id = commit_id; + match &mut self.task_type { + ArgsTaskType::Commit(task) | ArgsTaskType::CommitDiff(task) => { + task.commit_id = commit_id; + } + _ => {} + }
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
⛔ Files ignored due to path filters (2)
Cargo.lockis excluded by!**/*.locktest-integration/Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (28)
Cargo.toml(1 hunks)magicblock-accounts/src/scheduled_commits_processor.rs(3 hunks)magicblock-committor-service/src/tasks/args_task.rs(6 hunks)magicblock-committor-service/src/tasks/mod.rs(1 hunks)magicblock-committor-service/src/tasks/task_builder.rs(3 hunks)magicblock-committor-service/src/tasks/task_visitors/persistor_visitor.rs(1 hunks)magicblock-magic-program-api/src/instruction.rs(1 hunks)magicblock-rpc-client/src/lib.rs(1 hunks)programs/magicblock/src/magic_scheduled_base_intent.rs(7 hunks)programs/magicblock/src/magicblock_processor.rs(4 hunks)programs/magicblock/src/schedule_transactions/process_schedule_commit.rs(2 hunks)programs/magicblock/src/schedule_transactions/process_scheduled_commit_sent.rs(5 hunks)test-integration/Cargo.toml(2 hunks)test-integration/programs/schedulecommit-security/src/lib.rs(2 hunks)test-integration/programs/schedulecommit/src/api.rs(5 hunks)test-integration/programs/schedulecommit/src/lib.rs(9 hunks)test-integration/programs/schedulecommit/src/order_book.rs(1 hunks)test-integration/programs/schedulecommit/src/utils/mod.rs(3 hunks)test-integration/schedulecommit/client/src/schedule_commit_context.rs(6 hunks)test-integration/schedulecommit/client/src/verify.rs(2 hunks)test-integration/schedulecommit/test-scenarios/Cargo.toml(1 hunks)test-integration/schedulecommit/test-scenarios/tests/01_commits.rs(2 hunks)test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs(4 hunks)test-integration/schedulecommit/test-scenarios/tests/03_commits_fee_payer.rs(1 hunks)test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs(1 hunks)test-integration/test-ledger-restore/tests/08_commit_update.rs(2 hunks)test-integration/test-tools/src/integration_test_context.rs(2 hunks)test-integration/test-tools/src/scheduled_commits.rs(3 hunks)
🧰 Additional context used
🧠 Learnings (1)
📚 Learning: 2025-10-26T08:49:31.543Z
Learnt from: taco-paco
PR: magicblock-labs/magicblock-validator#585
File: magicblock-committor-service/src/tasks/buffer_task.rs:111-115
Timestamp: 2025-10-26T08:49:31.543Z
Learning: In the magicblock-committor-service, compute units returned by the `compute_units()` method in task implementations (such as `BufferTask`, `ArgsTask`, etc.) represent the compute budget for a single task. Transactions can comprise multiple tasks, and the total compute budget for a transaction is computed as the sum of the compute units of all tasks included in that transaction.
Applied to files:
magicblock-committor-service/src/tasks/args_task.rs
🧬 Code graph analysis (15)
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
test-integration/test-ledger-restore/src/lib.rs (1)
setup_validator_with_local_remote(97-119)
test-integration/schedulecommit/test-scenarios/tests/03_commits_fee_payer.rs (3)
test-integration/programs/schedulecommit/src/api.rs (1)
schedule_commit_with_payer_cpi_instruction(214-232)test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (3)
assert_two_committees_were_committed(68-91)get_context_with_delegated_committees(16-35)assert_feepayer_was_committed(94-113)test-integration/schedulecommit/client/src/verify.rs (1)
fetch_and_verify_commit_result_from_logs(7-14)
test-integration/programs/schedulecommit-security/src/lib.rs (1)
test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
create_schedule_commit_ix(41-66)
magicblock-committor-service/src/tasks/args_task.rs (4)
magicblock-committor-service/src/tasks/mod.rs (2)
instruction(66-66)instruction(249-255)magicblock-committor-service/src/tasks/buffer_task.rs (2)
instruction(65-86)new(37-45)magicblock-committor-service/src/config.rs (1)
local(29-35)magicblock-accounts/src/scheduled_commits_processor.rs (2)
new(66-90)new(420-436)
programs/magicblock/src/magicblock_processor.rs (3)
magicblock-committor-service/src/tasks/args_task.rs (1)
instruction(58-160)programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (1)
process_schedule_commit(34-259)magicblock-committor-program/src/state/changeset.rs (1)
request_undelegation(230-232)
test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs (4)
test-integration/programs/schedulecommit/src/api.rs (4)
schedule_commit_and_undelegate_cpi_instruction(234-252)schedule_commit_and_undelegate_cpi_with_mod_after_instruction(287-311)schedule_commit_diff_instruction_for_order_book(193-212)update_order_book_instruction(175-191)test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (3)
get_context_with_delegated_committees(16-35)assert_one_committee_was_committed(41-65)assert_one_committee_account_was_undelegated_on_chain(190-196)test-integration/schedulecommit/client/src/schedule_commit_context.rs (3)
committees(205-209)ephem_blockhash(322-324)ephem_client(318-320)test-integration/schedulecommit/client/src/verify.rs (1)
fetch_and_verify_order_book_commit_result_from_logs(16-23)
test-integration/programs/schedulecommit/src/api.rs (2)
test-integration/programs/schedulecommit/src/order_book.rs (1)
new(85-94)test-integration/programs/flexi-counter/src/state.rs (1)
pda(32-35)
test-integration/schedulecommit/client/src/verify.rs (2)
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
ctx(90-91)test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
ctx(145-146)
magicblock-committor-service/src/tasks/task_builder.rs (2)
magicblock-accounts/src/scheduled_commits_processor.rs (2)
new(66-90)new(420-436)magicblock-committor-service/src/tasks/args_task.rs (1)
new(49-54)
test-integration/programs/schedulecommit/src/lib.rs (2)
test-integration/programs/schedulecommit/src/utils/mod.rs (3)
assert_is_signer(30-44)assert_keys_equal(16-28)allocate_account_and_assign_owner(58-106)test-integration/programs/schedulecommit/src/order_book.rs (1)
new(85-94)
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (2)
magicblock-committor-service/src/tasks/task_builder.rs (1)
committed_accounts(153-156)test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
base_intent(544-549)
test-integration/schedulecommit/client/src/schedule_commit_context.rs (1)
test-integration/programs/schedulecommit/src/api.rs (3)
init_order_book_instruction(36-54)init_payer_escrow(77-98)init_account_instruction(16-34)
programs/magicblock/src/magic_scheduled_base_intent.rs (1)
magicblock-committor-service/src/tasks/task_builder.rs (1)
committed_accounts(153-156)
test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (3)
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
ctx(90-91)test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
ctx(145-146)test-integration/schedulecommit/client/src/schedule_commit_context.rs (3)
try_new(72-74)ncommittees(102-121)try_new_random_keys(66-71)
test-integration/schedulecommit/test-scenarios/tests/01_commits.rs (1)
test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (1)
get_context_with_delegated_committees(16-35)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (4)
- GitHub Check: run_make_ci_format
- GitHub Check: run_make_ci_test
- GitHub Check: run_make_ci_lint
- GitHub Check: run_make_ci_test
🔇 Additional comments (34)
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
55-55: LGTM! Correctly updated to match new function signature.Both call sites properly destructure the new 3-tuple return value. The
_tmpdirbinding is necessary to keep theTempDiralive throughout the function scope, preventing premature cleanup of the temporary directory while the validator is still using it.Also applies to: 170-170
test-integration/programs/schedulecommit/src/utils/mod.rs (2)
53-53: LGTM: Type alignment with Solana API.Changing
sizetou64aligns with Solana'ssystem_instruction::allocateAPI, which expects au64parameter. This improves type consistency and eliminates the need for type conversions downstream.
70-90: LGTM: Clean implementation with helpful debugging.The changes correctly handle the type conversion:
- Line 71: The cast to
usizeis necessary forRent::minimum_balanceand is safe given Solana's practical account size limits.- Lines 75-80: The debug log provides useful visibility into lamport requirements during account allocation.
- Line 90: Passing
sizedirectly asu64tosystem_instruction::allocateis cleaner and aligns with the API signature.test-integration/programs/schedulecommit-security/src/lib.rs (2)
2-2: LGTM: Import addition is correct.The
CommitPolicyimport is necessary for the API change on line 149.
143-150: No action required—the two CPI paths use intentionally different APIs.The review comment's concern about consistency is based on a false premise. The code uses two fundamentally different functions from different modules:
- Path 1 (
schedule_commit_cpi_instruction): A local program API wrapper with a fixed, simplified interface- Path 2 (
create_schedule_commit_ix): An external SDK function with extended configuration options includingCommitPolicyThese are not parallel implementations requiring symmetry. The
schedule_commit_cpi_instructionfunction does not supportCommitPolicyby design—it wraps commit scheduling with hardcoded policy defaults. AddingCommitPolicy::UseFullBytestocreate_schedule_commit_ixis the correct change and requires no corresponding update toschedule_commit_cpi_instruction.Likely an incorrect or invalid review comment.
test-integration/test-tools/src/integration_test_context.rs (1)
167-171: Transaction version configuration is correct and environment-aware.The conditional logic properly handles two distinct RPC endpoints:
- Chain (Solana devnet):
max_supported_transaction_version: Some(0)with 50 retries to handle devnet's known quirk of sometimes returning version responses instead of transaction metadata (noted at line 160-161).- Ephemeral (local validator):
Noneto use default behavior on a controlled, local environment without such quirks.Git history confirms this change is part of the current PR (commit 4c75d78). The code is correctly differentiated for each environment. No changes needed.
test-integration/test-tools/src/scheduled_commits.rs (1)
182-213: LGTM! Clear hierarchical naming for log variables.The renaming from
ephem_logs/chain_logstoephem_logs_l1/ephem_logs_l2improves clarity by establishing a consistent naming convention that reflects the two-level log retrieval hierarchy.magicblock-committor-service/src/tasks/task_visitors/persistor_visitor.rs (1)
29-33: LGTM: CommitDiff included in strategy persistence.Cleanly broadens Commit to CommitDiff without changing strategy semantics.
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (2)
224-230: Diff-aware commit routing is correct; keep logs concise.Choosing StandaloneDiff when request_diff is true is right. Consider demoting the "StandaloneDiff" ic_msg to debug in production builds to reduce log noise.
Also applies to: 233-238
29-32: All call sites properly initialize request_diff—no corrections needed.Verification confirms that every
process_schedule_commit()invocation explicitly sets bothrequest_undelegationandrequest_difffields. All three instruction paths (ScheduleCommit,ScheduleCommitAndUndelegate,ScheduleCommitDiffAndUndelegate) pass values forrequest_diff, withScheduleCommitDiffAndUndelegatecorrectly setting it totrueto trigger theCommitType::StandaloneDiffpath. No uninitialized defaults are used, and no silent routing to full-state commits can occur.programs/magicblock/src/schedule_transactions/process_scheduled_commit_sent.rs (1)
29-30: LGTM: commit_diff added end-to-end with safe logging.Field is carried into printable form and logged only as a flag. Tests initialize the new field.
Also applies to: 44-45, 71-73, 215-217, 254-255
magicblock-committor-service/src/tasks/args_task.rs (1)
197-205: Calibrate compute units for CommitDiff based on diff size.Fixed 65_000 may be inaccurate. Once diff is precomputed, scale CU by diff length/segments or add headroom to avoid CU errors.
Based on learnings
programs/magicblock/src/magicblock_processor.rs (3)
2-2: LGTM: import added is correct.
51-54: ScheduleCommit: explicit request_diff=false is correct.
90-99: New ScheduleCommitDiffAndUndelegate path correctly flips request_diff=true.magicblock-committor-service/src/tasks/task_builder.rs (1)
96-105: Per-account Commit vs CommitDiff task selection is correct.programs/magicblock/src/magic_scheduled_base_intent.rs (6)
104-107: ScheduledBaseIntent::is_commit_diff added — OK.
155-164: MagicBaseIntent::is_commit_diff added — OK.
448-454: CommitType::is_commit_diff implementation — OK.
456-463: Getters updated for StandaloneDiff — OK.Also applies to: 465-473
475-487: is_empty handles StandaloneDiff — OK.
317-323: StandaloneDiff variant already exists and is actively used in the codebase.The review's concern assumes
StandaloneDiffis a new addition that will shift bincode discriminants. However, the variant is already present inprocess_schedule_commit.rs(lines 225–226), indicating it exists before this PR or is not new.If
StandaloneDiffwas already committed to the codebase:
- The bincode discriminant shift already happened (or never was a concern).
- The review comment's warning is outdated or misguided.
If the snippet shows the final state (post-changes) and
StandaloneDiffwas genuinely added in this PR:
- Verify the diff shows it was inserted between existing variants (which would shift discriminants).
- If it was appended after existing variants, discriminants remain stable and no migration is needed.
Critical finding: No backwards compatibility tests, no versioning code, and no serde external tagging exist. If persisted state truly exists and the variant order changed, this is a risk. However, the evidence suggests
StandaloneDiffis not new.Action for developer: Confirm in the PR diff that you are not reordering existing enum variants. If appending new variants after existing ones, bincode compatibility is safe. If inserting between existing variants, migration is required.
test-integration/schedulecommit/test-scenarios/Cargo.toml (1)
19-20: LGTM!The addition of
randandborshas workspace dev-dependencies appropriately supports the new order-book test scenarios introduced in this PR.test-integration/schedulecommit/test-scenarios/tests/01_commits.rs (1)
30-31: LGTM!The addition of deterministic seed
b"magic_schedule_commit"improves test reproducibility by ensuring consistent PDA derivation across test runs.test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (2)
16-35: LGTM!The seed parameter enables deterministic test context initialization, and explicitly driving the
init_committeesanddelegate_committeesworkflow steps with debug logging improves test clarity and observability.
41-47: LGTM!Making
assert_one_committee_was_committedgeneric over typeTwith appropriate trait bounds enables reuse across different account types (MainAccount, OrderBookOwned, etc.) while maintaining type safety.test-integration/schedulecommit/test-scenarios/tests/03_commits_fee_payer.rs (2)
23-80: LGTM!The test properly validates that attempting to commit a fee payer without escrowing lamports fails with the expected "DoesNotHaveEscrowAccount" error. The test structure and assertions are appropriate.
83-135: LGTM!The test correctly validates the happy path where fee payer commits succeed when lamports are properly escrowed. The verification flow and assertions comprehensively check that both committees and the fee payer were committed and synchronized.
test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs (3)
55-56: LGTM!The addition of deterministic seed
b"magic_schedule_commit"ensures reproducible test behavior across runs.
110-164: LGTM!The
commit_and_undelegate_order_book_accounthelper follows the established pattern of other commit helpers in this file, properly constructing the update and schedule-commit-diff instructions sequence.
244-312: Excellent reproducibility design for randomized testing.The test demonstrates best practices for randomized testing:
- Prints the RNG seed for reproducibility
- Includes the seed in assertion failure messages
- Uses seeded RNG (StdRng) for deterministic behavior given a seed
This allows failures to be reproduced by rerunning with the printed seed value.
One optional enhancement: consider adding a way to override the seed via environment variable for easier failure reproduction:
let rng_seed = std::env::var("TEST_RNG_SEED") .ok() .and_then(|s| s.parse::<u64>().ok()) .unwrap_or_else(|| OsRng.next_u64());test-integration/programs/schedulecommit/src/lib.rs (1)
561-571: Use commit_diff for undelegate path; this looks correct.The diff-based commit is invoked when
undelegateis true; good alignment with the new optimization.Please confirm tests cover both diff and full-commit paths (with/without
undelegate).test-integration/programs/schedulecommit/src/order_book.rs (1)
101-116: Bounds/capacity logic is good; but relies on header being valid.Once header init is fixed in
process_init_order_book, these helpers are fine for basic growth semantics.Ensure tests cover:
- Non-zero existing asks/bids followed by additional inserts (to exercise prefix/suffix behavior).
- Capacity exhaustion returning
None.Also applies to: 118-138, 163-184
test-integration/schedulecommit/client/src/schedule_commit_context.rs (1)
260-267: Seed semantics are correct and consistently applied across the codebase.The verification shows that
delegate_account_cpi_instruction()explicitly branches on two supported canonical seeds:
b"magic_schedule_commit"→ScheduleCommitInstruction::DelegateCpib"order_book"→ScheduleCommitInstruction::DelegateOrderBookAll test utilities and callers supply only these supported seeds. The
delegate_committees()function at lines 260-267 correctly passes&self.user_seed, and all constructors receive only canonical seeds from test callers. PDA derivation and instruction creation use consistent seed semantics.
test-integration/schedulecommit/client/src/schedule_commit_context.rs
Outdated
Show resolved
Hide resolved
snawaz
force-pushed
the
snawaz/commit-diff
branch
2 times, most recently
from
9c44347 to
693af7f
Compare
There was a problem hiding this comment.
Actionable comments posted: 9
♻️ Duplicate comments (22)
magicblock-rpc-client/src/lib.rs (1)
430-430: Redundant error prefix persists in log message.The
error!macro already includes ERROR level in its output. The explicit"> ERROR:"prefix creates redundancy and reduces log readability.Apply this diff to remove the redundant prefix:
- error!("> ERROR: {:?}", err); + error!("{:?}", err);Or add context instead:
- error!("> ERROR: {:?}", err); + error!("Transaction processed status failed: {:?}", err);test-integration/test-tools/src/scheduled_commits.rs (2)
198-210: Verbose debug output concern still applies.The past review comment regarding verbose debug output from
println!with{:#?}formatting remains applicable to these new debug statements. Consider making the debug output conditional or using a more compact format.
233-235: Unnecessary clone() on Copy type.The past review comment regarding the redundant
.clone()call onSignature(which implementsCopy) remains applicable. Use*siginstead ofsig.clone().test-integration/test-tools/src/integration_test_context.rs (1)
157-158: Format the panic message withlabel.The panic still prints the literal
rpc_client for [{}] does not exist, so the label never appears when this trips. Please injectlabelinto the message.- let rpc_client = - rpc_client.expect("rpc_client for [{}] does not exist"); + let rpc_client = rpc_client.unwrap_or_else(|| { + panic!("rpc_client for [{}] does not exist", label) + });test-integration/schedulecommit/client/src/verify.rs (1)
16-23: Code duplication: consider the generic implementation suggested in the previous review.This function duplicates
fetch_and_verify_commit_result_from_logsexcept for the type parameter. A generic implementation withT: borsh::BorshDeserializewould eliminate this duplication.test-integration/schedulecommit/client/src/schedule_commit_context.rs (2)
115-118: PDA derivation must match program-side seeds.The PDA is derived using an arbitrary
user_seed, which may not match the on-chain program's expected seeds (e.g.,[b"order_book", book_manager]or[b"magic_schedule_commit", player]). This mismatch will cause transaction failures.
170-203: Validate user_seed explicitly to prevent silent misconfiguration.The implicit
elsebranch will execute for anyuser_seedthat isn'tb"magic_schedule_commit", including invalid values. Add explicit validation to fail fast with a clear error when an unsupported seed is provided.test-integration/programs/schedulecommit/src/api.rs (1)
106-109: Derive delegated PDA with canonical seeds to match on-chain expectations.The PDA is derived using an arbitrary
user_seed, but the on-chain program expects fixed seeds like[b"magic_schedule_commit", player]or[b"order_book", book_manager]. This mismatch will cause thedelegated_accountto not match program-side expectations.Cargo.toml (1)
112-112: CI/CD will fail: local path dependency not available.The path
../delegation-programdoes not exist in the repository or CI environments, causing all builds to fail. Either:
- Revert to a git-based dependency (consistent with the learning about using git branch references)
- Make
delegation-programa workspace member at the correct path- Update CI workflows to checkout the dependency
Verify the intended dependency resolution strategy:
#!/bin/bash # Check if the delegation-program path exists if [ -d "../delegation-program" ]; then echo "✓ Path exists locally" ls -la ../delegation-program/ else echo "✗ Path ../delegation-program does not exist" fi # Check for any git submodule configuration echo "" echo "=== Git submodules ===" git submodule status 2>/dev/null || echo "No submodules configured" # Check recent commits for this dependency change echo "" echo "=== Recent changes to delegation-program dependency ===" git log -n 5 --oneline --all -- Cargo.toml | head -10test-integration/Cargo.toml (1)
40-40: Critical: Local path dependencies will break CI builds.This issue has already been flagged in previous review comments. The local paths
../../ephemeral-rollups-sdk/rust/sdkand../../delegation-programdo not exist in the CI environment and will cause build failures.Also applies to: 60-60
test-integration/programs/schedulecommit/src/lib.rs (4)
302-309: Zero the order-book header immediately after allocation.
OrderBook::newreadsbids_len/asks_lenstraight from account data. Leaving those bytes uninitialized means random counts, which makes the next borrow calculate bogus slices and can walk past the buffer. Please clear the header bytes right after allocation (e.g. borrow data, zero the firstsize_of::<OrderBookHeader>()bytes).Apply this diff:
allocate_account_and_assign_owner(AllocateAndAssignAccountArgs { payer_info: payer, account_info: order_book, owner: &crate::ID, signer_seeds: &[b"order_book", book_manager.key.as_ref(), &[bump]], size: 10 * 1024, })?; + { + let mut data = order_book.try_borrow_mut_data()?; + let header_len = core::mem::size_of::<OrderBookHeader>(); + if data.len() < header_len { + return Err(ProgramError::AccountDataTooSmall); + } + for byte in &mut data[..header_len] { + *byte = 0; + } + } + Ok(())
329-334: Fix the PDA assertion message.The diagnostic still prints
payer.key, so a failing check points at the wrong seed. Swap it forbook_manager.keyas previously requested.Apply this diff:
assert_keys_equal(order_book.key, &pda, || { format!( "PDA for the account ('{}') and for book_manager ('{}') is incorrect", - order_book.key, payer.key + order_book.key, book_manager.key ) })?;
398-403: Update the log label.This path is mutating an existing order book, not initializing one. Please change the
msg!to say “Update order book” (or equivalent) so logs stay accurate.
424-451: Use the right error kind and require the payer to sign.When the accounts slice is the wrong length you must emit
ProgramError::NotEnoughAccountKeys;MissingRequiredSignatureis misleading. More importantly, we still never verifypayer.is_signer, so the CPI can run with an unsigned funder. Addassert_is_signer(payer, "payer")?before the CPI.Apply this diff:
let [payer, order_book_account, magic_context, magic_program] = accounts else { - return Err(ProgramError::MissingRequiredSignature); + return Err(ProgramError::NotEnoughAccountKeys); }; + assert_is_signer(payer, "payer")?; + commit_diff_and_undelegate_accounts( payer, vec![order_book_account], magic_context, magic_program,test-integration/programs/schedulecommit/src/order_book.rs (3)
45-55:deserializestill aliases immutable data mutably (UB).
slice::from_raw_parts_mutonbook_bytesforges a&mut [u8]from shared data, violating Rust’s aliasing rules and instantly triggering undefined behaviour. Please replace this with a safe path—e.g. copy into an owned buffer or implement a parser that reads header/levels straight from the immutable slice.
85-93:OrderBook::newassumes alignment that account data does not guarantee.Casting
header_bytes.as_ptr()to*mut OrderBookHeader(and the level pointer to*mut OrderLevel) requires 4/8‑byte alignment, but Solana account data is byte-aligned. The moment you dereference those pointers you trigger UB. Rework this to treat the buffer as raw bytes and read/write fields withfrom_le_bytes/to_le_bytesinstead of typed references.
140-178: Typed slices over account data remain unsafe.
slice::from_raw_parts(_ as *mut OrderLevel, ...)for bids/asks/buffer helpers creates slices ofOrderLevelout of unaligned memory. That’s the same UB called out earlier. Please keep the backing store as[u8], then decode/encode eachOrderLevelby copying 16 bytes into a local array and usingu64::from_le_bytes/u64::to_le_bytes.magicblock-magic-program-api/src/instruction.rs (1)
111-111: Add documentation for ScheduleCommitDiffAndUndelegate.This issue was already identified in a previous review. The variant lacks documentation while all sibling variants include doc comments describing the instruction purpose and account references.
magicblock-committor-service/src/tasks/args_task.rs (3)
75-103: Stop hard-coding the RPC endpoint in CommitDiff.
ChainConfig::local(ComputeBudgetConfig::new(1_000_000))forces every CommitDiff task to dialhttp://localhost:7799with Processed commitment and re-create an RPC client each timeinstruction()is called. In production that URI does not exist, so we immediately fall back tocommit_stateand never ship a diff; plus every call repeats the network fetch we already warned about. Inject the realChainConfig/RpcClientfrom the service (or precompute/cache the diff during task construction) soinstruction()stays pure and uses the configured endpoint. Prior feedback on this remains unresolved.
115-116: Do not log raw diff payloads.Dumping the entire diff at warn level leaks account data and explodes log volume. Log only bounded metadata (e.g., commit_id, pubkey, diff length) instead.
- log::warn!("DIFF computed: {:?}", args.diff); + log::debug!( + "commit_diff computed: commit_id={} pubkey={} len={}", + value.commit_id, + value.committed_account.pubkey, + args.diff.len() + );
171-173: Remove the CommitDiff panic in optimize().
optimize()is part of the normal pipeline; panicking here will abort the worker whenever a CommitDiff task flows through. ReturnErr(self)(mirroring the other variants) so the caller can fall back cleanly.- ArgsTaskType::CommitDiff(_) => { - panic!("ArgsTaskType::CommitDiff not handled") - } + ArgsTaskType::CommitDiff(_) => Err(self),programs/magicblock/src/magicblock_processor.rs (1)
38-38: Avoid logging entire MagicBlockInstruction payloads.
ic_msg!(..., "{:?}", instruction)dumps the full enum—including the diff bytes you just added—on every invocation. That inflates CU cost and spams logs, an issue we already called out earlier. Drop the log or gate it behind a debug-only feature flag so mainnet builds don’t emit the payload.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
⛔ Files ignored due to path filters (2)
Cargo.lockis excluded by!**/*.locktest-integration/Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (28)
Cargo.toml(1 hunks)magicblock-accounts/src/scheduled_commits_processor.rs(3 hunks)magicblock-committor-service/src/tasks/args_task.rs(6 hunks)magicblock-committor-service/src/tasks/mod.rs(1 hunks)magicblock-committor-service/src/tasks/task_builder.rs(3 hunks)magicblock-committor-service/src/tasks/task_visitors/persistor_visitor.rs(1 hunks)magicblock-magic-program-api/src/instruction.rs(1 hunks)magicblock-rpc-client/src/lib.rs(1 hunks)programs/magicblock/src/magic_scheduled_base_intent.rs(7 hunks)programs/magicblock/src/magicblock_processor.rs(4 hunks)programs/magicblock/src/schedule_transactions/process_schedule_commit.rs(2 hunks)programs/magicblock/src/schedule_transactions/process_scheduled_commit_sent.rs(5 hunks)test-integration/Cargo.toml(2 hunks)test-integration/programs/schedulecommit-security/src/lib.rs(2 hunks)test-integration/programs/schedulecommit/src/api.rs(5 hunks)test-integration/programs/schedulecommit/src/lib.rs(9 hunks)test-integration/programs/schedulecommit/src/order_book.rs(1 hunks)test-integration/programs/schedulecommit/src/utils/mod.rs(3 hunks)test-integration/schedulecommit/client/src/schedule_commit_context.rs(6 hunks)test-integration/schedulecommit/client/src/verify.rs(2 hunks)test-integration/schedulecommit/test-scenarios/Cargo.toml(1 hunks)test-integration/schedulecommit/test-scenarios/tests/01_commits.rs(2 hunks)test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs(4 hunks)test-integration/schedulecommit/test-scenarios/tests/03_commits_fee_payer.rs(1 hunks)test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs(1 hunks)test-integration/test-ledger-restore/tests/08_commit_update.rs(2 hunks)test-integration/test-tools/src/integration_test_context.rs(2 hunks)test-integration/test-tools/src/scheduled_commits.rs(3 hunks)
🧰 Additional context used
🧠 Learnings (4)
📚 Learning: 2025-10-14T09:56:14.047Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 564
File: test-integration/programs/flexi-counter/src/processor/call_handler.rs:122-125
Timestamp: 2025-10-14T09:56:14.047Z
Learning: The file test-integration/programs/flexi-counter/src/processor/call_handler.rs contains a test smart contract used for integration testing, not production code.
Applied to files:
test-integration/test-tools/src/integration_test_context.rstest-integration/programs/schedulecommit/src/utils/mod.rstest-integration/test-ledger-restore/tests/08_commit_update.rstest-integration/schedulecommit/test-scenarios/tests/03_commits_fee_payer.rs
📚 Learning: 2025-10-21T14:00:54.642Z
Learnt from: bmuddha
Repo: magicblock-labs/magicblock-validator PR: 578
File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27
Timestamp: 2025-10-21T14:00:54.642Z
Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579
Applied to files:
magicblock-committor-service/src/tasks/mod.rstest-integration/programs/schedulecommit/src/utils/mod.rsmagicblock-committor-service/src/tasks/args_task.rstest-integration/programs/schedulecommit/src/api.rstest-integration/programs/schedulecommit/src/lib.rstest-integration/programs/schedulecommit/src/order_book.rs
📚 Learning: 2025-10-26T16:54:39.084Z
Learnt from: thlorenz
Repo: magicblock-labs/magicblock-validator PR: 587
File: test-manual/Cargo.toml:0-0
Timestamp: 2025-10-26T16:54:39.084Z
Learning: In the magicblock-validator repository, use git branch references (not commit hashes or tags) for the helius-laserstream dependency to allow automatic updates when the branch is pushed to.
Applied to files:
test-integration/Cargo.tomlCargo.toml
📚 Learning: 2025-10-26T08:49:31.543Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 585
File: magicblock-committor-service/src/tasks/buffer_task.rs:111-115
Timestamp: 2025-10-26T08:49:31.543Z
Learning: In the magicblock-committor-service, compute units returned by the `compute_units()` method in task implementations (such as `BufferTask`, `ArgsTask`, etc.) represent the compute budget for a single task. Transactions can comprise multiple tasks, and the total compute budget for a transaction is computed as the sum of the compute units of all tasks included in that transaction.
Applied to files:
magicblock-committor-service/src/tasks/args_task.rs
🧬 Code graph analysis (15)
test-integration/programs/schedulecommit-security/src/lib.rs (1)
test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
create_schedule_commit_ix(41-66)
magicblock-committor-service/src/tasks/task_builder.rs (2)
magicblock-committor-service/src/tasks/args_task.rs (1)
new(49-54)magicblock-accounts/src/scheduled_commits_processor.rs (2)
new(66-90)new(420-436)
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (2)
magicblock-committor-service/src/tasks/task_builder.rs (1)
committed_accounts(153-156)test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
base_intent(544-549)
test-integration/schedulecommit/client/src/verify.rs (2)
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
ctx(90-91)test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
ctx(145-146)
programs/magicblock/src/magicblock_processor.rs (3)
magicblock-committor-service/src/tasks/args_task.rs (1)
instruction(58-160)programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (1)
process_schedule_commit(34-259)magicblock-committor-program/src/state/changeset.rs (1)
request_undelegation(230-232)
test-integration/schedulecommit/test-scenarios/tests/01_commits.rs (1)
test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (1)
get_context_with_delegated_committees(16-35)
magicblock-committor-service/src/tasks/args_task.rs (3)
magicblock-committor-service/src/tasks/mod.rs (2)
instruction(66-66)instruction(249-255)magicblock-committor-service/src/tasks/buffer_task.rs (1)
instruction(65-86)magicblock-committor-service/src/config.rs (1)
local(29-35)
test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (3)
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
ctx(90-91)test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
ctx(145-146)test-integration/schedulecommit/client/src/schedule_commit_context.rs (3)
try_new(72-74)ncommittees(102-121)try_new_random_keys(66-71)
programs/magicblock/src/magic_scheduled_base_intent.rs (1)
magicblock-committor-service/src/tasks/task_builder.rs (1)
committed_accounts(153-156)
test-integration/schedulecommit/client/src/schedule_commit_context.rs (2)
test-integration/programs/schedulecommit/src/api.rs (3)
init_order_book_instruction(36-54)init_payer_escrow(77-98)init_account_instruction(16-34)test-integration/test-tools/src/integration_test_context.rs (1)
try_new(113-115)
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
test-integration/test-ledger-restore/src/lib.rs (1)
setup_validator_with_local_remote(97-119)
test-integration/programs/schedulecommit/src/api.rs (2)
test-integration/programs/schedulecommit/src/order_book.rs (1)
new(85-94)test-integration/programs/flexi-counter/src/state.rs (1)
pda(32-35)
test-integration/schedulecommit/test-scenarios/tests/03_commits_fee_payer.rs (4)
test-integration/test-runner/bin/run_tests.rs (1)
run_test(777-796)test-integration/programs/schedulecommit/src/api.rs (1)
schedule_commit_with_payer_cpi_instruction(214-232)test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (4)
assert_two_committees_synchronized_count(139-169)assert_two_committees_were_committed(68-91)get_context_with_delegated_committees(16-35)assert_feepayer_was_committed(94-113)test-integration/schedulecommit/client/src/verify.rs (1)
fetch_and_verify_commit_result_from_logs(7-14)
test-integration/programs/schedulecommit/src/lib.rs (2)
test-integration/programs/schedulecommit/src/utils/mod.rs (3)
assert_is_signer(30-44)assert_keys_equal(16-28)allocate_account_and_assign_owner(58-106)test-integration/programs/schedulecommit/src/order_book.rs (1)
new(85-94)
test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs (4)
test-integration/programs/schedulecommit/src/api.rs (3)
schedule_commit_and_undelegate_cpi_with_mod_after_instruction(287-311)schedule_commit_diff_instruction_for_order_book(193-212)update_order_book_instruction(175-191)test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (3)
get_context_with_delegated_committees(16-35)assert_one_committee_was_committed(41-65)assert_one_committee_account_was_undelegated_on_chain(190-196)test-integration/schedulecommit/client/src/schedule_commit_context.rs (3)
committees(205-209)ephem_blockhash(322-324)ephem_client(318-320)test-integration/schedulecommit/client/src/verify.rs (1)
fetch_and_verify_order_book_commit_result_from_logs(16-23)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: run_make_ci_test
- GitHub Check: run_make_ci_lint
- GitHub Check: run_make_ci_test
🔇 Additional comments (20)
test-integration/programs/schedulecommit/src/utils/mod.rs (3)
49-55: LGTM: Type alignment with Solana API.Changing
sizefromusizetou64correctly aligns with Solana'ssystem_instruction::allocatesignature, which expectsspace: u64.
75-79: LGTM: Helpful debug logging.The added logging for lamport requirements and payer balance aids in troubleshooting account allocation issues during testing.
90-90: Critical fix: Correct system_instruction::allocate signature.The updated call now correctly provides both required arguments (
pubkeyandspace) tosystem_instruction::allocate. The previous implementation appears to have been missing the pubkey argument, which would not match the function signature:pub fn allocate(pubkey: &Pubkey, space: u64) -> Instruction.test-integration/test-tools/src/scheduled_commits.rs (1)
182-213: Variable renaming improves clarity.The renaming of log variables to
ephem_logs_l1andephem_logs_l2clearly distinguishes the two-level log fetching process, and the updated error context at line 206 now correctly references the first-level logs. These changes enhance code readability.test-integration/programs/schedulecommit-security/src/lib.rs (2)
2-2: LGTM!The
CommitPolicyimport is correctly added to support the new argument at line 149.
143-150: Ignore this review comment. The original comment incorrectly assumes thatschedule_commit_cpi_instructionandcreate_schedule_commit_ixshould be consistent, but they are two distinct functions from different crates with different purposes:
schedule_commit_cpi_instruction(fromprogram_schedulecommit::api) is a local CPI helper that does not accept aCommitPolicyparametercreate_schedule_commit_ix(fromephemeral_rollups_sdk::ephem) is an SDK function that acceptsCommitPolicyas a parameterThe two functions have intentionally different signatures. There is no test inconsistency to address.
Likely an incorrect or invalid review comment.
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
55-61: No issues found—code is correct as written.The
TempDirreturned bysetup_validator_with_local_remoteis created specifically for temporary config storage during validator initialization (written instart_magicblock_validator_with_config_structat line 225:fs::write(&config_path, config_toml)). After the validator process is spawned, this directory is no longer needed and is safely dropped. The actual ledger directory is managed separately by the main_tmpdirat line 42 of the test, which persists for the entire test duration. The binding to_tmpdirwith an underscore prefix correctly signals that this return value is intentionally unused beyond validator initialization.test-integration/programs/schedulecommit/src/api.rs (3)
36-54: LGTM!The
init_order_book_instructionfollows the standard instruction builder pattern with appropriate account metadata.
56-75: LGTM!The
grow_order_book_instructioncorrectly constructs the instruction with theadditional_spaceparameter.
175-212: LGTM!Both
update_order_book_instructionandschedule_commit_diff_instruction_for_order_bookfollow correct instruction builder patterns with appropriate account metadata.test-integration/schedulecommit/test-scenarios/tests/01_commits.rs (1)
30-31: LGTM! Deterministic seed ensures reproducible test contexts.The hard-coded seed
b"magic_schedule_commit"provides deterministic PDA derivation and context initialization, which improves test reliability and debugging.Also applies to: 84-85
test-integration/schedulecommit/test-scenarios/tests/03_commits_fee_payer.rs (2)
82-135: LGTM! Fee payer escrow test properly verifies expected behavior.The test correctly verifies that with proper escrow setup, the fee payer commit succeeds and all expected assertions pass.
34-41: Fix struct field destructuring:payer_ephemfield andephem_blockhashretrieval.The code attempts to destructure non-existent struct fields:
- Line 35:
payerfield doesn't exist inScheduleCommitTestContextFields— should bepayer_ephem: payer- Line 39:
ephem_blockhashis not a struct field — must be fetched separately viaephem_client.get_latest_blockhash().unwrap()- Lines 58, 112: Remove dereference (
*) fromephem_blockhashonce correctedSee
01_commits.rs(lines 34, 54) for the correct pattern.Also applies to: 88-95, 112
⛔ Skipped due to learnings
Learnt from: bmuddha Repo: magicblock-labs/magicblock-validator PR: 578 File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27 Timestamp: 2025-10-21T14:00:54.642Z Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579programs/magicblock/src/magic_scheduled_base_intent.rs (3)
104-106: LGTM!The
is_commit_diff()method correctly delegates to the base intent, following the same pattern as the existingis_undelegate()method.
155-163: LGTM!The
is_commit_diff()method correctly handles allMagicBaseIntentvariants, properly delegating to the underlyingCommitTypefor bothCommitandCommitAndUndelegatecases.
458-458: LGTM!The accessor methods correctly handle the new
StandaloneDiffvariant, treating it identically toStandalonesince both containVec<CommittedAccount>.Also applies to: 468-468, 480-482
magicblock-committor-service/src/tasks/task_visitors/persistor_visitor.rs (2)
29-33: LGTM!The match correctly handles both
CommitandCommitDiffvariants identically, as both containCommitTaskand should be persisted the same way.
59-59: ****The review comment is incorrect.
BufferTaskTypeintentionally has only a singleCommitvariant because it represents buffer-based commit operations.CommitDiffis a separate task variant that belongs exclusively toArgsTaskTypeand is handled through a different code path.Evidence:
- ArgsTaskType includes CommitDiff(CommitTask), but BufferTaskType does not
- ArgsTaskType::CommitDiff explicitly panics with "ArgsTaskType::CommitDiff not handled" when attempting conversion, indicating CommitDiff is intentionally not converted to BufferTask
- The refutable pattern at line 59 is safe because BufferTaskType can only ever contain a single variant
No code changes are needed; the refutable pattern is correct as implemented.
Likely an incorrect or invalid review comment.
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (2)
31-31: LGTM!The
request_difffield addition cleanly extends the options structure to control diff-based commit scheduling.
231-238: LGTM!The
commit_actionis correctly used in both the undelegation and non-undelegation paths, enabling diff-based commits for both scenarios while preserving existing semantics.
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs
Outdated
Show resolved
Hide resolved
test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs
Show resolved
Hide resolved
test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs
Show resolved
Hide resolved
test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs
Outdated
Show resolved
Hide resolved
snawaz
force-pushed
the
snawaz/commit-diff
branch
from
693af7f to
99f2e8f
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (8)
test-integration/Cargo.toml (1)
40-40:⚠️ Path dependency concerns from prior review remain unresolved.The migration from git-based to local path dependencies for
ephemeral-rollups-sdkandmagicblock-delegation-programstill poses the same critical CI/CD risks flagged in the previous review: these paths do not exist in the repository, CI workflows do not clone them, and builds will fail in CI/CD and for contributors without manual setup.The prior review's recommendations still apply:
- Update CI to clone both external repositories, OR
- Switch back to git-based references, OR
- Add clear setup documentation with exact clone commands and relative paths
Also applies to: 60-60
magicblock-committor-service/src/tasks/args_task.rs (2)
228-234:reset_commit_idmust handleCommitDiffvariant.The current implementation only updates the
commit_idforCommittasks and silently returns forCommitDiff. When aCommitDifftask is recycled, the stalecommit_idwill remain, causing desynced submissions.Apply this fix to handle both variants:
fn reset_commit_id(&mut self, commit_id: u64) { - // TODO (snawaz): handle CommitDiff as well? what is it about? - let ArgsTaskType::Commit(commit_task) = &mut self.task_type else { - return; - }; - - commit_task.commit_id = commit_id; + match &mut self.task_type { + ArgsTaskType::Commit(task) | ArgsTaskType::CommitDiff(task) => { + task.commit_id = commit_id; + } + _ => { + // Only Commit and CommitDiff tasks have commit_id + } + } }
74-122: Network I/O insideinstruction()remains a major concern.The
instruction()method performs RPC calls and diff computation on every invocation, which is expensive and risky:
instruction()may be called multiple times (e.g., viainvolved_accounts), multiplying RPC calls- TOCTOU: base-chain account can change between diff calculation and transaction submission
- Coupling to network hinders testing and determinism
Per previous discussion, this is acknowledged as temporary and will be addressed in subsequent PRs. Consider tracking this technical debt.
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (1)
224-228: Consider enhancing observability as previously suggested.The conditional logic correctly selects the commit type based on
request_diff. However, a past review comment suggests improving the log messages to include the number of committed accounts and the flag value for better operational visibility.test-integration/programs/schedulecommit/src/order_book.rs (2)
99-119: Unaligned typed reference toOrderBookHeaderremains a critical issue.Line 114 creates a typed reference
&mut OrderBookHeadervia raw pointer cast. While alignment is asserted at runtime (lines 102-110), Solana account data does not guarantee alignment for arbitrary types. This is undefined behavior if the account data is not aligned toalign_of::<OrderBookHeader>().The assertion will catch misalignment at runtime, but the proper fix is to avoid typed references entirely. Consider reading/writing the header fields as raw bytes using little-endian encoding.
165-178: Typed slices over potentially unaligned memory (OrderLevel).Lines 166 and 173-174 use
slice::from_raw_partsto create typed slices ofOrderLevel(which containsu64fields requiring 8-byte alignment). Account data alignment is not guaranteed by Solana, leading to undefined behavior on read.Per previous review: store levels as raw bytes and provide accessors that read/write fields via
u64::from_le_bytes/to_le_bytes, or use a packed representation without taking references to fields.test-integration/programs/schedulecommit/src/lib.rs (2)
280-311: Critical: OrderBook header is never initialized.After
allocate_account_and_assign_ownerreturns (line 308), the account data contains uninitialized memory. Subsequent calls toOrderBook::newwill read uninitializedbids_lenandasks_lenvalues, causing undefined behavior.Apply this fix immediately after allocation:
allocate_account_and_assign_owner(AllocateAndAssignAccountArgs { payer_info: payer, account_info: order_book, owner: &crate::ID, signer_seeds: &[b"order_book", book_manager.key.as_ref(), &[bump]], size: 10 * 1024, })?; + // Initialize header to zero + let mut data = order_book.try_borrow_mut_data()?; + let header_size = core::mem::size_of::<OrderBookHeader>(); + for b in &mut data[..header_size] { + *b = 0; + } Ok(())
329-334: Incorrect identifier in PDA assertion error message.Line 332 incorrectly references
payer.keyin the error message, which should bebook_manager.keyto accurately reflect the PDA derivation.Apply this fix:
assert_keys_equal(order_book.key, &pda, || { format!( "PDA for the account ('{}') and for book_manager ('{}') is incorrect", - order_book.key, payer.key + order_book.key, book_manager.key ) })?;
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
⛔ Files ignored due to path filters (2)
Cargo.lockis excluded by!**/*.locktest-integration/Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (8)
magicblock-committor-service/src/tasks/args_task.rs(6 hunks)programs/magicblock/src/magic_scheduled_base_intent.rs(7 hunks)programs/magicblock/src/magicblock_processor.rs(2 hunks)programs/magicblock/src/schedule_transactions/process_schedule_commit.rs(2 hunks)test-integration/Cargo.toml(4 hunks)test-integration/programs/schedulecommit/Cargo.toml(1 hunks)test-integration/programs/schedulecommit/src/lib.rs(9 hunks)test-integration/programs/schedulecommit/src/order_book.rs(1 hunks)
🧰 Additional context used
🧠 Learnings (2)
📚 Learning: 2025-10-21T14:00:54.642Z
Learnt from: bmuddha
Repo: magicblock-labs/magicblock-validator PR: 578
File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27
Timestamp: 2025-10-21T14:00:54.642Z
Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579
Applied to files:
magicblock-committor-service/src/tasks/args_task.rstest-integration/programs/schedulecommit/src/order_book.rstest-integration/programs/schedulecommit/src/lib.rs
📚 Learning: 2025-10-26T08:49:31.543Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 585
File: magicblock-committor-service/src/tasks/buffer_task.rs:111-115
Timestamp: 2025-10-26T08:49:31.543Z
Learning: In the magicblock-committor-service, compute units returned by the `compute_units()` method in task implementations (such as `BufferTask`, `ArgsTask`, etc.) represent the compute budget for a single task. Transactions can comprise multiple tasks, and the total compute budget for a transaction is computed as the sum of the compute units of all tasks included in that transaction.
Applied to files:
magicblock-committor-service/src/tasks/args_task.rs
🧬 Code graph analysis (5)
magicblock-committor-service/src/tasks/args_task.rs (3)
magicblock-committor-service/src/tasks/mod.rs (2)
instruction(66-66)instruction(249-255)magicblock-committor-service/src/tasks/buffer_task.rs (1)
instruction(65-86)magicblock-committor-service/src/config.rs (1)
local(29-35)
programs/magicblock/src/magicblock_processor.rs (1)
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (1)
process_schedule_commit(34-258)
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (2)
magicblock-committor-service/src/tasks/task_builder.rs (1)
committed_accounts(153-156)test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
base_intent(544-549)
programs/magicblock/src/magic_scheduled_base_intent.rs (1)
magicblock-committor-service/src/tasks/task_builder.rs (1)
committed_accounts(153-156)
test-integration/programs/schedulecommit/src/lib.rs (2)
test-integration/programs/schedulecommit/src/utils/mod.rs (3)
assert_is_signer(30-44)assert_keys_equal(16-28)allocate_account_and_assign_owner(58-106)test-integration/programs/schedulecommit/src/order_book.rs (1)
new(99-119)
🔇 Additional comments (11)
test-integration/Cargo.toml (1)
72-72: Now let me check the codebase for actual usage of these dependencies to confirm they are necessary:Based on my verification, I can now generate the rewritten review comment:
Both dependencies are compatible with Solana 2.2 and appropriate for this PR.
rkyv 0.7.45 supports no_std and no_alloc configurations, making it suitable for Solana on-chain programs. It supports no-std builds for targets without a standard library and no-alloc builds for targets where allocations cannot be made. The Solana VM guarantees that account data is byte-aligned to a multiple of 8, which enables rkyv's zero-copy deserialization to function safely on account data. static_assertions 1.1.0 is a stable, widely-adopted crate (225,914,390 downloads all time) providing compile-time assertion macros useful for validating account structure layouts and sizes—a common pattern in Solana programs.
No further verification needed.
test-integration/programs/schedulecommit/Cargo.toml (1)
11-12: LGTM! Dependencies properly support order book implementation.The additions of
rkyvandstatic_assertionsare well-utilized inorder_book.rsfor alignment guarantees and compile-time size/alignment checks.programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (1)
31-31: LGTM! New field enables diff-based commit path.The
request_diffflag properly controls whether to useCommitType::StandaloneDifforCommitType::Standalone, enabling the optimization for large accounts with small changes.programs/magicblock/src/magicblock_processor.rs (2)
44-59: LGTM! Existing instructions properly default to non-diff path.Both
ScheduleCommitandScheduleCommitAndUndelegatecorrectly setrequest_diff: falseto maintain existing behavior.
85-92: LGTM! New instruction variant correctly enables diff-based commit with undelegation.The
ScheduleCommitDiffAndUndelegateinstruction properly sets bothrequest_undelegation: trueandrequest_diff: true, mirroringScheduleCommitAndUndelegatebehavior while enabling the diff optimization.test-integration/programs/schedulecommit/src/order_book.rs (1)
48-75: LGTM! Mutable aliasing UB resolved with aligned copy.The implementation now safely copies the input buffer into
rkyv::AlignedVecbefore creating mutable references, avoiding the undefined behavior from the previous approach. The comment clearly explains the rationale.magicblock-committor-service/src/tasks/args_task.rs (2)
171-172: LGTM! CommitDiff correctly returnsErr(self)in optimize().Returning
Err(self)forCommitDiffis correct since the diff is already optimized by design—it sends only the differences rather than full account data. Converting toBufferTaskwould be counterproductive.Per previous discussion, this is the correct semantic since CommitDiff is already using the most efficient data transfer strategy.
196-203: LGTM! Compute units appropriately set for CommitDiff.The 65,000 compute units for
CommitDiff(vs 70,000 forCommit) reflects the reduced work from transmitting only the diff. This aligns with the test observation showing lower CU usage.Based on learnings: these are compute units per task, and transactions may comprise multiple tasks.
test-integration/programs/schedulecommit/src/lib.rs (3)
203-203: LGTM! Type change tou64is appropriate.Changing
MainAccount::SIZEfromusizetou64aligns with thesizeparameter type inallocate_account_and_assign_owner(line 255) and Solana's account size conventions.
398-414: LGTM! Log message and signer check are correct.Line 402 now uses the accurate log message "Update orderbook", and line 407 properly validates that the payer is a signer before allowing the update.
419-439: LGTM! Handler correctly uses diff-based commit with undelegation.The implementation properly:
- Returns
NotEnoughAccountKeyswhen accounts are missing (line 426)- Validates payer is a signer (line 429)
- Invokes
commit_diff_and_undelegate_accountsfor the optimized diff path (line 431)
snawaz
force-pushed
the
snawaz/commit-diff
branch
from
99f2e8f to
7eaaa13
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
magicblock-committor-service/src/tasks/args_task.rs (1)
228-235: Critical:reset_commit_idsilently ignores CommitDiff tasks.When a
CommitDifftask is recycled, the current code silently returns without updating itscommit_id(line 230-232), leaving stale IDs that will cause:
- Desynchronization with actual commit state
- Failed submissions with wrong nonce
- Data integrity issues
A past review correctly identified this issue.
Apply this fix to handle both
CommitandCommitDiff:fn reset_commit_id(&mut self, commit_id: u64) { - // TODO (snawaz): handle CommitDiff as well? what is it about? - let ArgsTaskType::Commit(commit_task) = &mut self.task_type else { - return; - }; - - commit_task.commit_id = commit_id; + match &mut self.task_type { + ArgsTaskType::Commit(task) | ArgsTaskType::CommitDiff(task) => { + task.commit_id = commit_id; + } + _ => { + // Other task types don't have commit_id + } + } }
♻️ Duplicate comments (7)
programs/magicblock/src/magic_scheduled_base_intent.rs (1)
389-446: CRITICAL: try_from_args cannot construct StandaloneDiff.The
try_from_argsmethod only handlesCommitTypeArgs::StandaloneandCommitTypeArgs::WithBaseActions, but does not handleCommitTypeArgs::StandaloneDiff. This means:
- Code paths using
try_from_args(e.g.,process_schedule_base_intent) cannot constructCommitType::StandaloneDiff- Only
process_schedule_commitcan create it programmatically- Any attempt to deserialize
CommitTypeArgs::StandaloneDifffrom instruction data will failThis creates a critical inconsistency where the
StandaloneDiffvariant cannot be fully utilized across all code paths.The method needs a new match arm similar to:
pub fn try_from_args( args: CommitTypeArgs, context: &ConstructionContext<'_, '_>, ) -> Result<CommitType, InstructionError> { match args { CommitTypeArgs::Standalone(accounts) => { // ... existing code ... Ok(CommitType::Standalone(committed_accounts)) } + CommitTypeArgs::StandaloneDiff(accounts) => { + let committed_accounts_ref = Self::extract_commit_accounts( + &accounts, + context.transaction_context, + )?; + Self::validate_accounts(&committed_accounts_ref, context)?; + let committed_accounts = committed_accounts_ref + .into_iter() + .map(|el| { + let mut committed_account: CommittedAccount = el.into(); + committed_account.account.owner = context + .parent_program_id + .unwrap_or(committed_account.account.owner); + committed_account + }) + .collect(); + Ok(CommitType::StandaloneDiff(committed_accounts)) + } CommitTypeArgs::WithBaseActions { ... } => { // ... existing code ... } } }Based on past review comment.
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (1)
224-228: Add observability logging for commit type selection.The conditional logic correctly selects the commit type, but lacks logging to indicate which path was taken. This makes debugging and auditing more difficult.
Based on past review comment, consider adding:
let commit_action = if opts.request_diff { + ic_msg!( + invoke_context, + "ScheduleCommit: using StandaloneDiff for {} accounts", + committed_accounts.len() + ); CommitType::StandaloneDiff(committed_accounts) } else { + ic_msg!( + invoke_context, + "ScheduleCommit: using Standalone for {} accounts", + committed_accounts.len() + ); CommitType::Standalone(committed_accounts) };magicblock-committor-service/src/tasks/args_task.rs (1)
74-122: Network I/O in instruction() confirmed as unresolved; improve fallback logging.The RPC-in-instruction issue is genuine and remains present despite the past review claiming it was addressed in commit 99f2e8f (which does not exist in the repository).
Confirmed concerns:
instruction() called multiple times per task: The method is invoked from
delivery_preparator.rs(line 377, mapped over cleanup_tasks),tasks/utils.rs(line 58, mapped over tasks), andtask_strategist.rs(line 217, in optimization loop). Each call triggers RPC I/O on the base-chain, multiplying network overhead and creating TOCTOU race conditions.ChainConfig hardcoded locally (line 76): ChainConfig is properly injected into the service/processor but ArgsTask hardcodes
ChainConfig::local()instead of receiving injected config. This remains unresolved.Fallback log incomplete (line 90): The message contains a typo ("commmit_id" with 3 m's) and is missing the
pubkeycontext for debugging. Improve clarity:- log::warn!("Fallback to commit_state and send full-bytes, as rpc failed to fetch the delegated-account from base chain, commmit_id: {} , error: {}", value.commit_id, e); + log::warn!( + "Fallback to commit_state for account {}: RPC fetch failed, commit_id={} - {}", + value.committed_account.pubkey, + value.commit_id, + e + );The architectural concerns about performing RPC I/O during instruction building should be addressed, and the config injection pattern needs to be established.
test-integration/programs/schedulecommit/src/lib.rs (3)
280-311: Header remains uninitialized after allocation—subsequent reads are undefined.After
allocate_account_and_assign_owner, the header fields (bids_len,asks_len) are still uninitialized. Any call toOrderBook::newon this account will read garbage.Apply this fix to zero the header immediately after allocation:
allocate_account_and_assign_owner(AllocateAndAssignAccountArgs { payer_info: payer, account_info: order_book, owner: &crate::ID, signer_seeds: &[b"order_book", book_manager.key.as_ref(), &[bump]], size: 10 * 1024, })?; + // Initialize header to zero + let mut data = order_book.try_borrow_mut_data()?; + data[..core::mem::size_of::<OrderBookHeader>()].fill(0); Ok(())
329-334: Error message references wrong key.The PDA assertion message incorrectly shows
payer.keyinstead ofbook_manager.key, making debugging harder.assert_keys_equal(order_book.key, &pda, || { format!( "PDA for the account ('{}') and for book_manager ('{}') is incorrect", - order_book.key, payer.key + order_book.key, book_manager.key ) })?;
402-402: Inconsistent log message casing.Message says "Update orderbook" but other logs use "OrderBook" with capital letters and space for consistency.
- msg!("Update orderbook"); + msg!("Update order book");test-integration/Cargo.toml (1)
40-40: Local path dependencies will break CI/CD builds.These local path dependencies point to repositories outside the current codebase (
../../ephemeral-rollups-sdkand../../delegation-program) that are not cloned by the CI workflow. This will cause build failures for all contributors and in continuous integration.Choose one of the following solutions:
- Revert to git-based dependencies (recommended for now):
-ephemeral-rollups-sdk = { path = "../../ephemeral-rollups-sdk/rust/sdk"} +ephemeral-rollups-sdk = { git = "https://github.com/magicblock-labs/ephemeral-rollups-sdk.git", rev = "..." }
Update CI to clone required repositories before build steps in
.github/workflows/ci-test-integration.ymlDocument the local setup in README with explicit clone instructions if local paths are intended for development only
Based on learnings: This appears to be a transition to local development setup but lacks the corresponding CI infrastructure updates.
Also applies to: 60-60
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
⛔ Files ignored due to path filters (2)
Cargo.lockis excluded by!**/*.locktest-integration/Cargo.lockis excluded by!**/*.lock
📒 Files selected for processing (8)
magicblock-committor-service/src/tasks/args_task.rs(6 hunks)programs/magicblock/src/magic_scheduled_base_intent.rs(7 hunks)programs/magicblock/src/magicblock_processor.rs(2 hunks)programs/magicblock/src/schedule_transactions/process_schedule_commit.rs(2 hunks)test-integration/Cargo.toml(4 hunks)test-integration/programs/schedulecommit/Cargo.toml(1 hunks)test-integration/programs/schedulecommit/src/lib.rs(9 hunks)test-integration/programs/schedulecommit/src/order_book.rs(1 hunks)
🧰 Additional context used
🧠 Learnings (2)
📚 Learning: 2025-10-21T14:00:54.642Z
Learnt from: bmuddha
Repo: magicblock-labs/magicblock-validator PR: 578
File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27
Timestamp: 2025-10-21T14:00:54.642Z
Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579
Applied to files:
magicblock-committor-service/src/tasks/args_task.rstest-integration/programs/schedulecommit/src/lib.rstest-integration/programs/schedulecommit/src/order_book.rs
📚 Learning: 2025-10-26T08:49:31.543Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 585
File: magicblock-committor-service/src/tasks/buffer_task.rs:111-115
Timestamp: 2025-10-26T08:49:31.543Z
Learning: In the magicblock-committor-service, compute units returned by the `compute_units()` method in task implementations (such as `BufferTask`, `ArgsTask`, etc.) represent the compute budget for a single task. Transactions can comprise multiple tasks, and the total compute budget for a transaction is computed as the sum of the compute units of all tasks included in that transaction.
Applied to files:
magicblock-committor-service/src/tasks/args_task.rs
🧬 Code graph analysis (5)
magicblock-committor-service/src/tasks/args_task.rs (4)
magicblock-committor-service/src/tasks/mod.rs (2)
instruction(66-66)instruction(249-255)magicblock-committor-service/src/tasks/buffer_task.rs (2)
instruction(65-86)new(37-45)magicblock-committor-service/src/config.rs (1)
local(29-35)magicblock-accounts/src/scheduled_commits_processor.rs (2)
new(66-90)new(420-436)
programs/magicblock/src/magic_scheduled_base_intent.rs (1)
magicblock-committor-service/src/tasks/task_builder.rs (1)
committed_accounts(153-156)
programs/magicblock/src/magicblock_processor.rs (1)
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (1)
process_schedule_commit(34-258)
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (2)
magicblock-committor-service/src/tasks/task_builder.rs (1)
committed_accounts(153-156)test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
base_intent(544-549)
test-integration/programs/schedulecommit/src/lib.rs (2)
test-integration/programs/schedulecommit/src/utils/mod.rs (3)
assert_is_signer(30-44)assert_keys_equal(16-28)allocate_account_and_assign_owner(58-106)test-integration/programs/schedulecommit/src/order_book.rs (1)
new(105-125)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: run_make_ci_format
- GitHub Check: run_make_ci_test
- GitHub Check: run_make_ci_test
🔇 Additional comments (14)
programs/magicblock/src/schedule_transactions/process_schedule_commit.rs (1)
29-32: Field addition looks correct.The new
request_difffield is properly scoped aspub(crate)and theDefaulttrait will initialize it tofalse, maintaining backward compatibility.programs/magicblock/src/magicblock_processor.rs (2)
44-51: ScheduleCommit handler correctly configured.The handler properly sets
request_diff: falseto maintain the existing full-commit behavior for standard commits.
85-92: New instruction handler correctly implemented.The
ScheduleCommitDiffAndUndelegatehandler properly enables both undelegation and diff-based commits by setting both flags totrue. The implementation follows the established pattern from the other handlers.programs/magicblock/src/magic_scheduled_base_intent.rs (4)
104-106: Delegation method correctly implemented.The
is_commit_diff()method appropriately delegates to the underlyingbase_intent, following the established pattern for other query methods in this struct.
155-163: Commit diff detection logic is correct.The method correctly returns
falseforBaseActionsand properly delegates to the commit-related variants to determine if differential commits are enabled.
317-327: StandaloneDiff variant added correctly.The new variant follows the same structure as
Standalone, holdingVec<CommittedAccount>. Note that full account data is stored here; the actual diff computation likely occurs downstream during commit processing.
448-484: Accessor methods properly handle StandaloneDiff.All match expressions are now exhaustive and correctly handle the new
StandaloneDiffvariant. Theis_commit_diff()method correctly returnstrueonly forStandaloneDiff, and the accessor methods treat it consistently withStandalone.magicblock-committor-service/src/tasks/args_task.rs (4)
1-24: LGTM: Imports support the CommitDiff feature.All new imports are necessary for the RPC client setup, diff computation, and configuration.
30-30: LGTM: CommitDiff variant added.Reusing
CommitTaskfor theCommitDiffvariant is appropriate since they share the same data requirements.
171-172: LGTM: ReturningErr(self)for CommitDiff is correct.The past review discussion thoroughly analyzed this and concluded that
CommitDiffshould returnErr(self)(no optimization) because:
CommitDiffis already optimized by design—it transmits only the diff (e.g., 286 bytes) vs. full account data (10,240 bytes per your test)- Converting
CommitDifftoBufferTaskwould be counterproductive and defeat its purposeErr(self)correctly signals "no further optimization possible"Suggestion: The TODO on line 171 can be resolved and removed based on the past review's analysis.
199-199: LGTM: Compute units appropriately reduced for CommitDiff.The 65,000 compute unit allocation (5k less than
Commit's 70k) aligns with your test observations and reflects the reduced data processing for diff-based commits.test-integration/programs/schedulecommit/src/lib.rs (1)
419-439: LGTM—signer validation and error handling are now correct.The function now properly returns
NotEnoughAccountKeyswhen the accounts array doesn't match, and validates thatpayeris a signer before proceeding with the CPI call.test-integration/programs/schedulecommit/Cargo.toml (1)
11-12: LGTM—dependencies appropriately support zero-copy OrderBook.The
rkyvcrate providesAlignedVecfor safe buffer allocation, andstatic_assertionsenables compile-time layout verification. Both are correctly utilized inorder_book.rs.test-integration/Cargo.toml (1)
72-72: LGTM—workspace dependencies properly declared.The
rkyvandstatic_assertionscrates are correctly added at the workspace level, enabling their use across integration test crates for zero-copy structures and compile-time assertions.Also applies to: 91-91
snawaz
force-pushed
the
snawaz/commit-diff
branch
2 times, most recently
from
db80cab to
dcaafd1
Compare
There was a problem hiding this comment.
Actionable comments posted: 3
♻️ Duplicate comments (3)
test-integration/test-committor-service/tests/test_delivery_preparator.rs (1)
22-24: Remove temporary debug print statements.These
println!statements appear to be temporary debugging output and were previously flagged for removal. Consider removing them or replacing with proper test logging if diagnostics are needed.Apply this diff to remove the debug statements:
- println!("TestFixture::new()"); let fixture = TestFixture::new().await; - println!("TestFixture::new() done");magicblock-committor-service/src/tasks/mod.rs (2)
116-116: Document the rationale for COMMIT_STATE_SIZE_THRESHOLD.The 200-byte threshold determines when to use CommitDiff (RPC-based) vs CommitState (direct), but there's no explanation of:
- Why 200 bytes was chosen
- The trade-offs (RPC overhead vs instruction data size)
- Whether this should be configurable
Add a doc comment:
impl CommitTask { + /// Accounts larger than 200 bytes use CommitDiff to reduce instruction size. + /// Below 200 bytes, the RPC overhead of fetching base-chain state exceeds + /// the benefit of sending a diff instead of full state. + /// TODO: Consider making this configurable via config/env. const COMMIT_STATE_SIZE_THRESHOLD: usize = 200;Based on past review comments.
118-156: CRITICAL: Blocking RPC call with hard-coded localhost in constructor.This constructor has multiple critical issues that prevent CommitDiff from working in production:
Blocking network I/O in constructor: Constructors should not make network calls. This blocks the thread and violates async/sync boundaries. The PR notes mention tests now require
#[tokio::test(flavor = "multi_thread")]precisely because of this blocking call.Hard-coded localhost endpoint: Lines 133-141 create an RPC client pointing to
http://localhost:7799viaChainConfig::local. In any non-local environment (dev, staging, prod), this fetch will always fail, causing CommitDiff to never execute and always fall back to CommitState. This defeats the entire purpose of the PR.Silent error handling: Line 143 uses
.ok()which swallows all RPC errors without logging. When the fetch fails in production, there's no visibility into why.No caching: This expensive RPC call runs on every
CommitTask::new(), even for the same account.TOCTOU window: Fetching at construction time and using later widens the time-of-check-to-time-of-use gap.
The TODO at lines 126-127 acknowledges this is the "ugliest piece of code" but it's a blocker - the feature is non-functional outside localhost.
Required fixes:
Remove blocking call from constructor: Accept a pre-fetched
Option<Account>parameter instead, or make this an async constructor, or use a builder pattern with async fetch.Inject the configured RPC client: Accept
ChainConfigorArc<RpcClient>as a parameter so the real endpoint is used, not localhost.Add error logging: Log when RPC fetch fails so there's visibility in production.
Consider caching layer: As the TODO mentions, implement caching to avoid repeated fetches.
Example approach (inject config):
impl CommitTask { pub fn new( commit_id: u64, allow_undelegation: bool, committed_account: CommittedAccount, + chain_config: &ChainConfig, ) -> Self { let fetched_account = if committed_account.account.data.len() > CommitTask::COMMIT_STATE_SIZE_THRESHOLD { - use solana_rpc_client::rpc_client::RpcClient; - use solana_sdk::commitment_config::CommitmentConfig; - - use crate::{config::ChainConfig, ComputeBudgetConfig}; - - let chain_config = - ChainConfig::local(ComputeBudgetConfig::new(1_000_000)); - let rpc_client = RpcClient::new_with_commitment( chain_config.rpc_uri.to_string(), CommitmentConfig { commitment: chain_config.commitment, }, ); - rpc_client.get_account(&committed_account.pubkey).ok() + match rpc_client.get_account(&committed_account.pubkey) { + Ok(account) => Some(account), + Err(e) => { + log::warn!("Failed to fetch base account {} for commit_id {}: {}", + committed_account.pubkey, commit_id, e); + None + } + } } else { None };Or better yet, move the RPC fetch out of the constructor entirely and accept
base_account: Option<Account>as a parameter.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
📒 Files selected for processing (2)
magicblock-committor-service/src/tasks/mod.rs(10 hunks)test-integration/test-committor-service/tests/test_delivery_preparator.rs(4 hunks)
🧰 Additional context used
🧠 Learnings (2)
📚 Learning: 2025-10-14T09:56:14.047Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 564
File: test-integration/programs/flexi-counter/src/processor/call_handler.rs:122-125
Timestamp: 2025-10-14T09:56:14.047Z
Learning: The file test-integration/programs/flexi-counter/src/processor/call_handler.rs contains a test smart contract used for integration testing, not production code.
Applied to files:
test-integration/test-committor-service/tests/test_delivery_preparator.rs
📚 Learning: 2025-10-21T14:00:54.642Z
Learnt from: bmuddha
Repo: magicblock-labs/magicblock-validator PR: 578
File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27
Timestamp: 2025-10-21T14:00:54.642Z
Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579
Applied to files:
magicblock-committor-service/src/tasks/mod.rs
🧬 Code graph analysis (2)
test-integration/test-committor-service/tests/test_delivery_preparator.rs (2)
magicblock-committor-service/src/tasks/mod.rs (1)
new(118-156)test-integration/test-committor-service/tests/common.rs (1)
new(52-55)
magicblock-committor-service/src/tasks/mod.rs (3)
magicblock-committor-service/src/tasks/args_task.rs (1)
new(36-41)magicblock-committor-service/src/tasks/buffer_task.rs (1)
new(37-45)magicblock-committor-service/src/config.rs (1)
local(29-35)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: Build Project
- GitHub Check: run_make_ci_lint
- GitHub Check: run_make_ci_test
🔇 Additional comments (4)
test-integration/test-committor-service/tests/test_delivery_preparator.rs (4)
18-20: LGTM! Clear documentation of runtime requirement.The TODO comment properly explains the necessity for multi-threaded test execution due to blocking RpcClient calls in CommitTask::new(). The change is well-justified and aligns with the PR objectives.
83-85: LGTM! Consistent with multi-threaded runtime requirement.The TODO comment and multi-threaded test attribute are consistent with the blocking RpcClient issue documented in PR #575. The test creates large buffers (10KB, 500KB) that trigger the threshold requiring RPC prefetch.
164-164: Correctly remains single-threaded.This test appropriately uses the standard
#[tokio::test]attribute since it only creates small data (10, 20, 30 bytes), which is well below theCOMMIT_STATE_SIZE_THRESHOLDand won't trigger blocking RpcClient calls inCommitTask::new().
215-217: LGTM! Consistent multi-threaded runtime configuration.Both TODO comments and test attribute changes are appropriate and consistent with the pattern established throughout the file. The tests create large buffers (10KB+ and 12KB/64KB respectively) that exceed the threshold and require blocking RPC calls.
Also applies to: 299-301
snawaz
force-pushed
the
snawaz/commit-diff
branch
from
6f44e7f to
10f1f92
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
⛔ Files ignored due to path filters (1)
test-integration/schedulecommit/elfs/dlp.sois excluded by!**/*.so
📒 Files selected for processing (1)
test-integration/test-committor-service/tests/test_ix_commit_local.rs(1 hunks)
🧰 Additional context used
🧠 Learnings (1)
📚 Learning: 2025-10-14T09:56:14.047Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 564
File: test-integration/programs/flexi-counter/src/processor/call_handler.rs:122-125
Timestamp: 2025-10-14T09:56:14.047Z
Learning: The file test-integration/programs/flexi-counter/src/processor/call_handler.rs contains a test smart contract used for integration testing, not production code.
Applied to files:
test-integration/test-committor-service/tests/test_ix_commit_local.rs
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: Build Project
- GitHub Check: run_make_ci_lint
- GitHub Check: run_make_ci_test
🔇 Additional comments (1)
test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
216-218: AI summary incorrectly identifies the changed test.The AI summary states the change applies to
test_ix_commit_four_accounts_1kb_2kb_5kb_10kb_single_bundle, but the actual change is totest_commit_20_accounts_1kb_bundle_size_2(line 219).
snawaz
force-pushed
the
snawaz/commit-diff
branch
from
10f1f92 to
03efa01
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
216-218: Apply the multi-threaded annotation consistently to allcommit_20_accounts_1kbtests.This annotation was added to prevent runtime panics when
CommitTask::new()executes blocking RpcClient calls. However, as identified in the previous review, the following tests execute the identical code path and require the same annotation:
- Line 288:
test_commit_20_accounts_1kb_bundle_size_3- Line 296:
test_commit_20_accounts_1kb_bundle_size_4- Line 305:
test_commit_20_accounts_1kb_bundle_size_6- Line 319:
test_commit_20_accounts_1kb_bundle_size_20- Line 341:
test_commit_20_accounts_1kb_bundle_size_8The
bundle_sizeparameter only affects bundle creation logic (line 414) and does not bypass the blockingCommitTask::new()call.Apply this pattern to all affected tests:
+// TODO (snawaz): use #[tokio::test] once CommitTask::new() stops using blocking RpcClient +// # see the PR #575 for more context. +#[tokio::test(flavor = "multi_thread", worker_threads = 2)] -#[tokio::test] async fn test_commit_20_accounts_1kb_bundle_size_X() {
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
⛔ Files ignored due to path filters (1)
test-integration/schedulecommit/elfs/dlp.sois excluded by!**/*.so
📒 Files selected for processing (2)
test-integration/test-committor-service/tests/test_ix_commit_local.rs(2 hunks)test-integration/test-committor-service/tests/utils/transactions.rs(1 hunks)
🧰 Additional context used
🧠 Learnings (2)
📚 Learning: 2025-10-14T09:56:14.047Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 564
File: test-integration/programs/flexi-counter/src/processor/call_handler.rs:122-125
Timestamp: 2025-10-14T09:56:14.047Z
Learning: The file test-integration/programs/flexi-counter/src/processor/call_handler.rs contains a test smart contract used for integration testing, not production code.
Applied to files:
test-integration/test-committor-service/tests/test_ix_commit_local.rs
📚 Learning: 2025-10-21T14:00:54.642Z
Learnt from: bmuddha
Repo: magicblock-labs/magicblock-validator PR: 578
File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27
Timestamp: 2025-10-21T14:00:54.642Z
Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579
Applied to files:
test-integration/test-committor-service/tests/test_ix_commit_local.rs
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: Build Project
- GitHub Check: run_make_ci_lint
- GitHub Check: run_make_ci_test
🔇 Additional comments (1)
test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
150-151: LGTM!The TODO comment clearly documents future test coverage plans for larger commit sizes.
There was a problem hiding this comment.
Actionable comments posted: 1
♻️ Duplicate comments (1)
magicblock-committor-service/src/tasks/mod.rs (1)
124-149: Hard-coded localhost RPC disables CommitDiff outside dev boxes
CommitTask::newspins up anRpcClientwithChainConfig::local(...)(http://localhost:7799, Processed) and uses it to fetch the base account. In production the committor talks to the configured cluster RPC, so this call always fails, leavingbase_account = None. That forcescreate_commit_ixto fall back tocommit_state, so the CommitDiff path never executes anywhere except a localhost setup. It also re-opens the redundant client/TOCTOU problem raised earlier. Please inject the realChainConfig/RpcClientfrom the service layer and reuse it here instead of hard-codingChainConfig::local, so the diff fetch succeeds in real deployments.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
📒 Files selected for processing (2)
magicblock-committor-service/src/tasks/args_task.rs(3 hunks)magicblock-committor-service/src/tasks/mod.rs(10 hunks)
🧰 Additional context used
🧠 Learnings (2)
📚 Learning: 2025-10-21T14:00:54.642Z
Learnt from: bmuddha
Repo: magicblock-labs/magicblock-validator PR: 578
File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27
Timestamp: 2025-10-21T14:00:54.642Z
Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579
Applied to files:
magicblock-committor-service/src/tasks/args_task.rsmagicblock-committor-service/src/tasks/mod.rs
📚 Learning: 2025-10-14T09:56:14.047Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 564
File: test-integration/programs/flexi-counter/src/processor/call_handler.rs:122-125
Timestamp: 2025-10-14T09:56:14.047Z
Learning: The file test-integration/programs/flexi-counter/src/processor/call_handler.rs contains a test smart contract used for integration testing, not production code.
Applied to files:
magicblock-committor-service/src/tasks/args_task.rs
🧬 Code graph analysis (2)
magicblock-committor-service/src/tasks/args_task.rs (2)
magicblock-committor-service/src/tasks/mod.rs (1)
new(119-158)magicblock-committor-service/src/tasks/buffer_task.rs (2)
new(37-45)new_preparation_required(30-35)
magicblock-committor-service/src/tasks/mod.rs (1)
magicblock-committor-service/src/config.rs (1)
local(29-35)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: run_make_ci_test
- GitHub Check: Build Project
- GitHub Check: run_make_ci_lint
snawaz
force-pushed
the
snawaz/commit-diff
branch
from
2ff5502 to
fb096a7
Compare
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
173-246: Fix threshold inconsistency: single-account tests (800B, 1KB, 10KB) should expect Args, not FromBuffer.The COMMIT_STATE_SIZE_THRESHOLD is 200 bytes. CommitDiff (Args strategy) is used when account size > 200 bytes AND base_account is successfully fetched.
The multi-account tests correctly expect Args for account sizes > 200 bytes (lines 179–235). However, the single-account tests contradict this:
- Line 83: 800 bytes expects
FromBuffer→ should expectArgs(800 > 200)- Line 90: 800 bytes expects
FromBuffer→ should expectArgs- Line 97: 1KB expects
FromBuffer→ should expectArgs- Line 104: 10KB expects
FromBuffer→ should expectArgsUpdate these four tests to expect
CommitStrategy::Argsto align with the threshold logic and the multi-account test expectations.
♻️ Duplicate comments (5)
magicblock-committor-service/src/tasks/args_task.rs (1)
91-100: CRITICAL: CommitDiff optimization is completely disabled.This code defeats the entire CommitDiff feature that the PR is intended to add:
- When
is_commit_diff()returnstrue(large account with fetched base), this branch executesforce_commit_state()is called, setting a flag that makesis_commit_diff()returnfalse- The task is wrapped in a
BufferTask- When
BufferTasklater callscreate_commit_ix(), the forced flag causes fallback tocreate_commit_state_ix()- Result: Full account state is sent instead of the diff, nullifying the optimization
According to your test results,
CommitDiffshould senddiff len = 286bytes versusdata len = 10240bytes. But with this code, the 10KB full state will always be sent for large accounts.The strategist always calls
optimize(), so this branch fires 100% of the time for diff-capable tasks.Per the TODO comment, if
BufferTaskcannot supportCommitDiff, then returnErr(self)here (likeBaseAction,Finalize,Undelegatedo) so the task stays on the Args path and can actually submit the diff.Apply this fix:
- ArgsTaskType::Commit(mut value) if value.is_commit_diff() => { - //TODO (snawaz): We do not currently support executing CommitDiff as BufferTask, which is why we're forcing CommitTask to use CommitState before converting this task into BufferTask - // Once CommitDiff is supported by BufferTask, we do not have to - // force_commit_state. - - value.force_commit_state(); - Ok(Box::new(BufferTask::new_preparation_required( - BufferTaskType::Commit(value), - ))) - } + ArgsTaskType::Commit(value) if value.is_commit_diff() => { + // CommitDiff cannot be converted to BufferTask because the diff + // is computed against the current base-chain state and must be + // sent directly in instruction args. Return Err to keep on Args path. + Err(self) + }Based on past review comments.
magicblock-committor-service/src/tasks/mod.rs (4)
117-117: Document the rationale for COMMIT_STATE_SIZE_THRESHOLD.The 200-byte threshold determines when to attempt CommitDiff (with RPC fetch) versus direct CommitState. Adding a comment would help maintainers understand:
- Why 200 bytes was chosen
- The trade-offs (RPC overhead vs instruction data size)
- Whether this should be configurable
Example:
impl CommitTask { + // Accounts larger than 200 bytes may use CommitDiff to reduce instruction size. + // Below 200 bytes, the RPC overhead outweighs the benefit of sending a diff. + // TODO: Make this threshold configurable based on network conditions. const COMMIT_STATE_SIZE_THRESHOLD: usize = 200;Based on past review comments.
119-158: CRITICAL: Hard-coded localhost RPC prevents CommitDiff in production.The constructor creates an
RpcClientpointing toChainConfig::local()which hard-codes"http://localhost:7799"(seeconfig.rs:28-34). This means CommitDiff will NEVER work in any non-local environment:
- Production/staging validators don't run RPC on localhost:7799
get_account()will fail, settingbase_account = Noneis_commit_diff()returnsfalse, falling back to full state- Your test showing "diff len = 286" only works because it runs locally
Additional problems:
- Blocking RPC call in constructor blocks the tokio runtime
- No timeout configured—hangs indefinitely on slow/dead endpoints
- TOCTOU: base-chain account can change between fetch and submit
- Constructor doing network I/O is an anti-pattern
Solution: Inject a configured
RpcClient(orChainConfig) intoCommitTask::new()from the service layer that creates tasks. The service already has the real chain config—pass it through the task builder.pub fn new( commit_id: u64, allow_undelegation: bool, committed_account: CommittedAccount, + rpc_client: Option<&RpcClient>, ) -> Self { let fetched_account = if committed_account.account.data.len() > CommitTask::COMMIT_STATE_SIZE_THRESHOLD { - use solana_rpc_client::rpc_client::RpcClient; - use solana_sdk::commitment_config::CommitmentConfig; - - use crate::{config::ChainConfig, ComputeBudgetConfig}; - - let chain_config = - ChainConfig::local(ComputeBudgetConfig::new(1_000_000)); - - let rpc_client = RpcClient::new_with_commitment( - chain_config.rpc_uri.to_string(), - CommitmentConfig { - commitment: chain_config.commitment, - }, - ); - - rpc_client.get_account(&committed_account.pubkey).ok() + rpc_client.and_then(|client| { + client.get_account(&committed_account.pubkey).ok() + }) } else { None }; - - println!("fetched_account: {:#?}", fetched_account);Based on past review comments.
129-132: Move imports to module level.Imports inside the constructor (lines 129-131) are a code smell suggesting this code was added hastily. These should be at module level with other imports.
However, once you fix the critical RPC-in-constructor issue by injecting the client, these imports may no longer be needed here at all.
Based on past review comments.
149-149: Remove debug println! statements.Debug
println!at lines 149, 186, and 217 flood stdout and bypass configured log levels. Remove them before merge, or replace withlog::debug!if needed for troubleshooting.- println!("fetched_account: {:#?}", fetched_account);Also applies to: 186-186, 217-217
Based on past review comments.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
📒 Files selected for processing (3)
magicblock-committor-service/src/tasks/args_task.rs(3 hunks)magicblock-committor-service/src/tasks/mod.rs(10 hunks)test-integration/test-committor-service/tests/test_ix_commit_local.rs(10 hunks)
🧰 Additional context used
🧠 Learnings (2)
📚 Learning: 2025-10-14T09:56:14.047Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 564
File: test-integration/programs/flexi-counter/src/processor/call_handler.rs:122-125
Timestamp: 2025-10-14T09:56:14.047Z
Learning: The file test-integration/programs/flexi-counter/src/processor/call_handler.rs contains a test smart contract used for integration testing, not production code.
Applied to files:
test-integration/test-committor-service/tests/test_ix_commit_local.rsmagicblock-committor-service/src/tasks/args_task.rs
📚 Learning: 2025-10-21T14:00:54.642Z
Learnt from: bmuddha
Repo: magicblock-labs/magicblock-validator PR: 578
File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27
Timestamp: 2025-10-21T14:00:54.642Z
Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579
Applied to files:
test-integration/test-committor-service/tests/test_ix_commit_local.rsmagicblock-committor-service/src/tasks/mod.rsmagicblock-committor-service/src/tasks/args_task.rs
🧬 Code graph analysis (3)
test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
magicblock-validator/src/main.rs (1)
init_logger(13-50)
magicblock-committor-service/src/tasks/mod.rs (1)
magicblock-committor-service/src/config.rs (1)
local(29-35)
magicblock-committor-service/src/tasks/args_task.rs (2)
magicblock-committor-service/src/tasks/mod.rs (1)
new(119-158)magicblock-committor-service/src/tasks/buffer_task.rs (2)
new(37-45)new_preparation_required(30-35)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (3)
- GitHub Check: Build Project
- GitHub Check: run_make_ci_test
- GitHub Check: run_make_ci_lint
🔇 Additional comments (12)
magicblock-committor-service/src/tasks/args_task.rs (3)
1-1: LGTM: Import cleanup aligns with delegation pattern.Removing
CommitStateArgsfrom imports is correct since commit instruction creation is now delegated toCommitTask::create_commit_ix()at line 47.
47-47: LGTM: Proper delegation to CommitTask.Delegating instruction creation to
CommitTask::create_commit_ix()is the correct approach, allowing CommitTask to internally choose between CommitState and CommitDiff paths.
157-163: LGTM: reset_commit_id correctly handles both commit paths.This implementation is correct. Both
CommitStateandCommitDiffexecution paths use the sameArgsTaskType::Commit(CommitTask)variant, so this code properly resets the commit_id for all commit-related tasks.The distinction between diff and state is internal to
CommitTask(viais_commit_diff()), not a separate enum variant.magicblock-committor-service/src/tasks/mod.rs (6)
1-4: LGTM: Necessary imports for CommitDiff functionality.The new imports (
CommitDiffArgs,CommitStateArgs,compute_diff) are required to support both commit instruction variants.
20-20: LGTM: Required imports for account handling.
AccountandReadableAccountimports are necessary for thebase_accountfield and RPC operations.
160-169: LGTM: Helper methods correctly implement diff detection logic.The
is_commit_diff()andforce_commit_state()methods properly encapsulate the logic for determining when to use CommitDiff versus CommitState. The checks for threshold, base_account presence, and force flag are appropriate.
171-177: LGTM: Clean dispatch logic for commit instruction creation.The
create_commit_ix()method cleanly dispatches to the appropriate instruction builder based on whether a base account was fetched. This is the correct entry point for instruction generation.
195-225: LGTM: CommitDiff logic correctly computes and sends diff.The
create_commit_diff_ix()method properly:
- Calls
compute_diff()with base and current account data- Constructs
CommitDiffArgswith the computed diff- Uses the DLP instruction builder
The defensive check at lines 200-205 is redundant (since
create_commit_ix()already handles this), but it's not harmful and provides an extra safety layer.Note: The
println!at line 217 should be removed (already flagged separately).
432-446: LGTM: Test correctly updated to use new CommitTask constructor.The test properly uses
CommitTask::new()instead of struct literal initialization, aligning with the new API. The small account data (3 bytes) won't trigger CommitDiff path, which is appropriate for a serialization test.Also applies to: 492-506, 517-531
test-integration/test-committor-service/tests/test_ix_commit_local.rs (3)
64-106: Multi-threading annotations correctly applied.All single-account tests now properly use
#[tokio::test(flavor = "multi_thread", worker_threads = 2)]to avoid runtime panics from blockingRpcClientcalls inCommitTask::new(). This addresses the issue identified in previous reviews.The TODO comments, while repetitive, provide clear context for future refactoring when the blocking call is removed.
248-391: Comprehensive test coverage with consistent multi-threading annotations.All remaining tests properly use multi-threaded runtime annotations and cover various scenarios:
- Different bundle sizes (2, 3, 4, 5, 6, 8, 20 accounts per bundle)
- Multiple account counts (5, 8, 20 accounts total)
- Commit-only and commit-with-undelegate flows
The strategy expectations have been systematically updated to reflect CommitDiff behavior, maintaining consistency across the test suite.
569-575: The review comment is incorrect—the assertion already handles both CommitState and CommitDiff correctly.The
tx_logs_containutility function (intests/utils/transactions.rslines 124–129) explicitly handles both instruction types when checking for "CommitState". The function's logic checkslog.contains("CommitState") || log.contains("CommitDiff")when the needle is "CommitState". The comments in that function explain this was an intentional design to avoid requiring tests to compute which strategy was used and pass the correct instruction name.The assertion at line 570 is correct as-is and requires no changes.
Likely an incorrect or invalid review comment.
snawaz
force-pushed
the
snawaz/commit-diff
branch
from
fb096a7 to
588ec52
Compare
…fficiently commit changes in delegated accounts
snawaz
force-pushed
the
snawaz/commit-diff
branch
from
588ec52 to
36d77aa
Compare
There was a problem hiding this comment.
Actionable comments posted: 2
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
magicblock-committor-service/src/tasks/task_strategist.rs (1)
159-171: TODO flags valid performance concern.The TODO comment correctly identifies that the
calculate_tx_lengthclosure performs heavy computations that may be discarded during the optimization loop. Consider tracking which tasks have already been serialized/measured to avoid redundant work, or restructure the optimization to be more incremental.
♻️ Duplicate comments (16)
test-integration/test-tools/src/scheduled_commits.rs (1)
198-198: Consider verbosity of debug output.The
println!statements with{:#?}pretty-printing could produce very verbose output, especially if the log arrays are large. While acceptable for test tooling, this may clutter test output and make it harder to identify actual test failures.Consider either:
- Making these conditional via an environment variable (e.g.,
MAGICBLOCK_DEBUG_LOGS)- Using a more compact format like
{:?}, or- Confirming this is temporary debugging code to be removed before merge
Also applies to: 210-210
test-integration/schedulecommit/client/src/schedule_commit_context.rs (1)
166-169: Consider extracting compute budget values as named constants.The compute unit limit (1,400,000) and price (10,000) are hardcoded. Extracting these as named constants would improve maintainability.
Example:
const INIT_COMPUTE_UNIT_LIMIT: u32 = 1_400_000; const INIT_COMPUTE_UNIT_PRICE: u64 = 10_000; let mut ixs = vec![ ComputeBudgetInstruction::set_compute_unit_limit(INIT_COMPUTE_UNIT_LIMIT), ComputeBudgetInstruction::set_compute_unit_price(INIT_COMPUTE_UNIT_PRICE), ];test-integration/test-committor-service/tests/test_delivery_preparator.rs (3)
81-83: Same multi-threading workaround applied consistently.This test also requires the multi-threaded runtime due to the blocking RpcClient calls in
CommitTask::new().
213-215: Multi-threading workaround applied consistently.This test similarly requires the multi-threaded runtime due to blocking RpcClient calls.
297-299: Multi-threading workaround applied consistently.This test also requires the multi-threaded runtime for the same reason documented in earlier tests.
programs/magicblock/src/magic_scheduled_base_intent.rs (1)
374-431: Critical: MissingStandaloneDiffhandling intry_from_args.The PR objectives describe a new
CommitDiffflow, butCommitType::try_from_argsonly constructsStandaloneandWithBaseActionsvariants. According to past review comments, there should be aStandaloneDiffvariant that needs to be constructed fromCommitTypeArgs::StandaloneDiff. Without this, the new diff-based commit instruction cannot function—allis_commit_diff()calls will return false and the runtime will continue down the full-commit path.Please extend
try_from_argsto recognizeCommitTypeArgs::StandaloneDiffand construct the correspondingCommitType::StandaloneDiffvariant, including extracting and storing the diff bytes into theCommittedAccountrepresentation.test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
193-193: Strategy expectations consistently updated across tests.Multiple tests now expect
CommitStrategy::Argsinstead of buffer-based strategies, reflecting the new commit path that uses instruction args for smaller data payloads.Also applies to: 207-207, 221-221, 235-235, 244-245, 254-254, 269-270, 283-283, 298-299, 322-323, 346-346
test-integration/schedulecommit/test-scenarios/tests/01_commits.rs (1)
84-85: Consistent seed parameter usage.The second test also correctly passes the seed parameter for deterministic context creation.
magicblock-committor-service/src/tasks/args_task.rs (2)
159-165: reset_commit_id must handle CommitDiff.When a
CommitDifftask is recycled,reset_commit_idsilently returns without updating thecommit_id, leaving stale data. Extend the pattern match:- let ArgsTaskType::Commit(commit_task) = &mut self.task_type else { - return; - }; - - commit_task.commit_id = commit_id; + match &mut self.task_type { + ArgsTaskType::Commit(task) => { + task.commit_id = commit_id; + } + _ => {} + }
91-102: Critical: force_commit_state defeats CommitDiff optimization.Forcing every
CommitDifftask back toCommitStatebefore converting toBufferTaskcompletely nullifies the diff optimization. Since the strategist always callsoptimize(),CommitDiffinstructions never reach execution.The TODO acknowledges this is temporary, but it means the entire CommitDiff feature is non-functional in this PR. Consider:
- Return
Err(self)forCommitDiff(like other non-optimizable tasks) to preserve the diff path, OR- Clearly document in the PR description that CommitDiff is merged but disabled until BufferTask support is added
Your test shows
diff len = 286vsdata len = 10240, which is excellent—but this code path ensures production never uses it.test-integration/programs/schedulecommit/src/utils/mod.rs (1)
71-71: Consider checked conversion for the cast.The unchecked
as usizecast could theoretically truncate on 32-bit platforms ifsizeexceedsusize::MAX. While Solana programs run in 64-bit environments, a checked conversion (e.g.,size.try_into()?) would be more defensive.test-integration/schedulecommit/client/src/verify.rs (1)
16-23: Eliminate duplication by making the function generic.This function duplicates
fetch_and_verify_commit_result_from_logswith only the type parameter differing. Consider making the original function generic:-pub fn fetch_and_verify_commit_result_from_logs( +pub fn fetch_and_verify_commit_result_from_logs<T>( ctx: &ScheduleCommitTestContext, sig: Signature, -) -> ScheduledCommitResult<MainAccount> { +) -> ScheduledCommitResult<T> +where + T: borsh::BorshDeserialize, +{ let res = ctx.fetch_schedule_commit_result(sig).unwrap(); res.confirm_commit_transactions_on_chain(ctx).unwrap(); res } - -pub fn fetch_and_verify_order_book_commit_result_from_logs( - ctx: &ScheduleCommitTestContext, - sig: Signature, -) -> ScheduledCommitResult<OrderBookOwned> { - let res = ctx.fetch_schedule_commit_result(sig).unwrap(); - res.confirm_commit_transactions_on_chain(ctx).unwrap(); - res -}Call sites would use turbofish:
fetch_and_verify_commit_result_from_logs::<OrderBookOwned>(&ctx, sig).test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (1)
41-47: Remove unused trait bounds.The trait bounds
borsh::BorshDeserialize + PartialEq + Eq + std::fmt::Debugare not used by this function—the body only inspects theScheduledCommitResultstructure without deserializing or comparing values of typeT.pub fn assert_one_committee_was_committed<T>( ctx: &ScheduleCommitTestContext, res: &ScheduledCommitResult<T>, is_single_stage: bool, -) -where - T: std::fmt::Debug + borsh::BorshDeserialize + PartialEq + Eq, -{ +) {test-integration/programs/schedulecommit/src/lib.rs (1)
332-348: Guardnew_sizeagainst overflow before reallocating.
order_book.data_len() + additional_space as usizecan wrap on BPF. A largeadditional_spacewill shrink the buffer and truncate the order book while the instruction still succeeds.Use checked arithmetic (and bail on overflow) before calling
realloc:- let new_size = order_book.data_len() + additional_space as usize; + let additional = usize::try_from(additional_space) + .map_err(|_| ProgramError::InvalidArgument)?; + let new_size = order_book + .data_len() + .checked_add(additional) + .ok_or(ProgramError::InvalidArgument)?;This keeps the account safe even when callers request huge growth.
test-integration/programs/schedulecommit/src/order_book.rs (1)
127-130: Fail fast if the book cannot absorb an update.
add_bids/add_asksreturnOptionprecisely to signal capacity exhaustion, but the current code ignores it. When the book is full the instruction still returnsOk(()), leaving the caller with stale state and no warning.Propagate the failure (or at least panic) so capacity issues surface immediately:
- self.add_bids(&updates.bids); - self.add_asks(&updates.asks); + self.add_bids(&updates.bids) + .expect("order book capacity exhausted while applying bids"); + self.add_asks(&updates.asks) + .expect("order book capacity exhausted while applying asks");That way the transaction aborts instead of silently dropping updates.
magicblock-committor-service/src/tasks/mod.rs (1)
121-158: Remove the hard-coded localhost RPC — CommitDiff never fires outside tests.Every construction of
CommitTaskspins up a freshRpcClientpointed atChainConfig::local("http://localhost:7799"). In any non-local deployment that endpoint is unavailable, soget_accountfails,base_accountstaysNone, andcreate_commit_ixalways falls back tocommit_state. The CommitDiff path is therefore dead in production, defeating the feature this PR introduces. It also recreates a blocking client per constructor call, widening the TOCTOU window.Please inject the real, configured RPC (or a pre-fetched base account) from the caller instead of hard-coding localhost here—e.g. extend the constructor to accept an
&RpcClientor a trait-backed fetcher and reuse it:- pub fn new( - commit_id: u64, - allow_undelegation: bool, - committed_account: CommittedAccount, - ) -> Self { + pub fn new( + commit_id: u64, + allow_undelegation: bool, + committed_account: CommittedAccount, + rpc_client: &RpcClient, + ) -> Self { let fetched_account = if committed_account.account.data.len() > CommitTask::COMMIT_STATE_SIZE_THRESHOLD { - // TODO (snawaz) … - rpc_client.get_account(&committed_account.pubkey).ok() + rpc_client.get_account(&committed_account.pubkey).ok() } else { None };Thread the configured client (or the fetched
Account) through the task builder so CommitDiff can execute against the real endpoint.
📜 Review details
Configuration used: CodeRabbit UI
Review profile: ASSERTIVE
Plan: Pro
⛔ Files ignored due to path filters (3)
Cargo.lockis excluded by!**/*.locktest-integration/Cargo.lockis excluded by!**/*.locktest-integration/schedulecommit/elfs/dlp.sois excluded by!**/*.so
📒 Files selected for processing (27)
Cargo.toml(1 hunks)magicblock-committor-service/src/tasks/args_task.rs(3 hunks)magicblock-committor-service/src/tasks/mod.rs(10 hunks)magicblock-committor-service/src/tasks/task_builder.rs(1 hunks)magicblock-committor-service/src/tasks/task_strategist.rs(3 hunks)programs/magicblock/src/magic_scheduled_base_intent.rs(1 hunks)test-integration/Cargo.toml(4 hunks)test-integration/programs/schedulecommit/Cargo.toml(1 hunks)test-integration/programs/schedulecommit/src/api.rs(5 hunks)test-integration/programs/schedulecommit/src/lib.rs(8 hunks)test-integration/programs/schedulecommit/src/order_book.rs(1 hunks)test-integration/programs/schedulecommit/src/utils/mod.rs(3 hunks)test-integration/schedulecommit/client/src/schedule_commit_context.rs(6 hunks)test-integration/schedulecommit/client/src/verify.rs(2 hunks)test-integration/schedulecommit/test-scenarios/Cargo.toml(1 hunks)test-integration/schedulecommit/test-scenarios/tests/01_commits.rs(2 hunks)test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs(4 hunks)test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs(1 hunks)test-integration/schedulecommit/test-security/tests/01_invocations.rs(1 hunks)test-integration/test-committor-service/tests/common.rs(2 hunks)test-integration/test-committor-service/tests/test_delivery_preparator.rs(4 hunks)test-integration/test-committor-service/tests/test_ix_commit_local.rs(10 hunks)test-integration/test-committor-service/tests/test_transaction_preparator.rs(3 hunks)test-integration/test-committor-service/tests/utils/transactions.rs(1 hunks)test-integration/test-ledger-restore/tests/08_commit_update.rs(2 hunks)test-integration/test-tools/src/integration_test_context.rs(2 hunks)test-integration/test-tools/src/scheduled_commits.rs(3 hunks)
🧰 Additional context used
🧠 Learnings (3)
📚 Learning: 2025-10-14T09:56:14.047Z
Learnt from: taco-paco
Repo: magicblock-labs/magicblock-validator PR: 564
File: test-integration/programs/flexi-counter/src/processor/call_handler.rs:122-125
Timestamp: 2025-10-14T09:56:14.047Z
Learning: The file test-integration/programs/flexi-counter/src/processor/call_handler.rs contains a test smart contract used for integration testing, not production code.
Applied to files:
test-integration/test-committor-service/tests/test_ix_commit_local.rstest-integration/programs/schedulecommit/src/utils/mod.rstest-integration/test-ledger-restore/tests/08_commit_update.rstest-integration/test-tools/src/scheduled_commits.rstest-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rsmagicblock-committor-service/src/tasks/args_task.rstest-integration/test-tools/src/integration_test_context.rs
📚 Learning: 2025-10-21T14:00:54.642Z
Learnt from: bmuddha
Repo: magicblock-labs/magicblock-validator PR: 578
File: magicblock-aperture/src/requests/websocket/account_subscribe.rs:18-27
Timestamp: 2025-10-21T14:00:54.642Z
Learning: In magicblock-aperture account_subscribe handler (src/requests/websocket/account_subscribe.rs), the RpcAccountInfoConfig fields data_slice, commitment, and min_context_slot are currently ignored—only encoding is applied. This is tracked as technical debt in issue #579: https://github.com/magicblock-labs/magicblock-validator/issues/579
Applied to files:
test-integration/test-committor-service/tests/test_ix_commit_local.rstest-integration/programs/schedulecommit/src/utils/mod.rstest-integration/programs/schedulecommit/src/api.rsmagicblock-committor-service/src/tasks/args_task.rsprograms/magicblock/src/magic_scheduled_base_intent.rstest-integration/programs/schedulecommit/src/order_book.rstest-integration/programs/schedulecommit/src/lib.rstest-integration/schedulecommit/client/src/schedule_commit_context.rsmagicblock-committor-service/src/tasks/mod.rs
📚 Learning: 2025-10-26T16:54:39.084Z
Learnt from: thlorenz
Repo: magicblock-labs/magicblock-validator PR: 587
File: test-manual/Cargo.toml:0-0
Timestamp: 2025-10-26T16:54:39.084Z
Learning: In the magicblock-validator repository, use git branch references (not commit hashes or tags) for the helius-laserstream dependency to allow automatic updates when the branch is pushed to.
Applied to files:
test-integration/Cargo.tomlCargo.toml
🧬 Code graph analysis (16)
test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
test-integration/schedulecommit/client/src/schedule_commit_context.rs (2)
try_new(72-74)try_new_random_keys(66-71)
magicblock-committor-service/src/tasks/task_builder.rs (2)
magicblock-committor-service/src/tasks/args_task.rs (1)
new(36-41)magicblock-committor-service/src/tasks/mod.rs (1)
new(121-159)
test-integration/test-committor-service/tests/test_ix_commit_local.rs (1)
magicblock-validator/src/main.rs (1)
init_logger(13-50)
test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (3)
test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
ctx(148-149)test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
ctx(90-91)test-integration/schedulecommit/client/src/schedule_commit_context.rs (3)
try_new(72-74)ncommittees(102-121)try_new_random_keys(66-71)
test-integration/schedulecommit/client/src/verify.rs (2)
test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
ctx(148-149)test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
ctx(90-91)
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
test-integration/test-ledger-restore/src/lib.rs (1)
setup_validator_with_local_remote(97-119)
magicblock-committor-service/src/tasks/task_strategist.rs (2)
magicblock-committor-service/src/tasks/args_task.rs (1)
new(36-41)magicblock-committor-service/src/tasks/mod.rs (1)
new(121-159)
test-integration/programs/schedulecommit/src/api.rs (1)
test-integration/programs/schedulecommit/src/order_book.rs (1)
new(105-125)
test-integration/schedulecommit/test-scenarios/tests/01_commits.rs (1)
test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (1)
get_context_with_delegated_committees(16-35)
test-integration/test-committor-service/tests/common.rs (1)
magicblock-committor-service/src/tasks/mod.rs (1)
new(121-159)
test-integration/schedulecommit/test-scenarios/tests/02_commit_and_undelegate.rs (4)
test-integration/programs/schedulecommit/src/api.rs (3)
schedule_commit_and_undelegate_cpi_with_mod_after_instruction(287-311)schedule_commit_diff_instruction_for_order_book(193-212)update_order_book_instruction(175-191)test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (1)
get_context_with_delegated_committees(16-35)test-integration/programs/schedulecommit/src/order_book.rs (1)
bids(171-173)test-integration/schedulecommit/client/src/verify.rs (1)
fetch_and_verify_order_book_commit_result_from_logs(16-23)
magicblock-committor-service/src/tasks/args_task.rs (2)
magicblock-committor-service/src/tasks/mod.rs (1)
new(121-159)magicblock-committor-service/src/tasks/buffer_task.rs (2)
new(37-45)new_preparation_required(30-35)
test-integration/test-committor-service/tests/test_transaction_preparator.rs (2)
magicblock-committor-service/src/tasks/mod.rs (1)
new(121-159)magicblock-committor-service/src/tasks/buffer_task.rs (2)
new(37-45)new_preparation_required(30-35)
test-integration/programs/schedulecommit/src/lib.rs (2)
test-integration/programs/schedulecommit/src/utils/mod.rs (3)
assert_is_signer(30-44)assert_keys_equal(16-28)allocate_account_and_assign_owner(58-106)test-integration/programs/schedulecommit/src/order_book.rs (1)
new(105-125)
test-integration/schedulecommit/client/src/schedule_commit_context.rs (1)
test-integration/programs/schedulecommit/src/api.rs (3)
init_order_book_instruction(36-54)init_payer_escrow(77-98)init_account_instruction(16-34)
magicblock-committor-service/src/tasks/mod.rs (1)
magicblock-committor-service/src/config.rs (1)
local(29-35)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Build Project
🔇 Additional comments (30)
test-integration/schedulecommit/test-security/tests/01_invocations.rs (1)
28-42: LGTM! Seed parameter correctly threaded through test context initialization.The addition of the
user_seedparameter (b"magic_schedule_commit") to bothtry_new()andtry_new_random_keys()calls is consistent and aligns with the deterministic seed mechanism introduced across the test harness. The static seed is appropriate for test code and ensures reproducibility.test-integration/test-tools/src/integration_test_context.rs (2)
157-159: LGTM! Panic message fix is correct.The format placeholder issue has been properly resolved. The panic message now correctly interpolates the
labelvariable, so errors will clearly indicate which RPC client ("chain" or "ephemeral") is missing.
168-174: LGTM! Conditional logic is now documented.The comment clearly explains why different clusters use different transaction version settings. The implementation is appropriate for test infrastructure where the chain cluster needs explicit v0 transaction support.
test-integration/test-ledger-restore/tests/08_commit_update.rs (1)
55-61: LGTM! Correct handling of the temporary directory.Both call sites properly capture the
TempDirreturned bysetup_validator_with_local_remote. Binding it to_tmpdirensures the temporary directory stays alive for the correct scope (entire function duration), preventing premature cleanup that could affect validator operation.Also applies to: 170-176
test-integration/test-tools/src/scheduled_commits.rs (3)
182-196: LGTM! Clear naming for two-stage log handling.The variable rename to
ephem_logs_l1clearly distinguishes the first ephemeral log fetch from the subsequent second fetch, improving code readability for the new two-stage flow.
201-209: LGTM! Second-stage log fetch properly implemented.The second ephemeral log fetch using
scheduled_commmit_sent_sigis correctly implemented with appropriate error context referencing the first-stage logs. This two-stage pattern aligns with the newCommitDiffflow described in the PR objectives.
212-214: LGTM! Correctly uses second-stage logs for extraction.The extraction now properly consumes
ephem_logs_l2to get the commit information, which is the expected behavior in the two-stage log handling flow.test-integration/schedulecommit/client/src/schedule_commit_context.rs (6)
8-8: LGTM: Imports aligned with new functionality.The new imports support the order book initialization path and compute budget configuration.
Also applies to: 16-16
34-34: LGTM: user_seed properly encapsulated.The private field and
.to_vec()storage are appropriate.Also applies to: 158-158
66-80: LGTM: Constructor API updated consistently.All constructors now require
user_seed, maintaining a consistent API surface.
170-192: LGTM: Seed-based initialization branching is clean.The match on
user_seedcorrectly dispatches toinit_account_instructionorinit_order_book_instruction.
271-276: LGTM: user_seed correctly propagated to delegation call.Passing
&self.user_seedtodelegate_account_cpi_instructionaligns with the updated API.
115-118: PDA derivation is correct for both cases; no changes needed.The client derives PDAs at lines 115-118 using
[user_seed, payer_ephem.pubkey()]. Verification confirms:
For
b"order_book": Client passespayer_ephemasbook_managerand the derived PDA. On-chainprocess_init_order_bookderives[b"order_book", book_manager.key](line 287) and validates it matches the passed account (line 291). ✓For
b"magic_schedule_commit": Client passespayer_ephemasplayerand the derived PDA. On-chainprocess_initderives[b"magic_schedule_commit", player_info.key](line 221), which matches. ✓Both derivations are consistent with the on-chain handlers.
test-integration/schedulecommit/test-scenarios/Cargo.toml (1)
19-20: LGTM!Adding
randandborshas dev-dependencies is appropriate for enhanced test data generation and serialization support in the schedule commit test scenarios.test-integration/test-committor-service/tests/test_delivery_preparator.rs (1)
18-20: Consistent workaround for blocking RpcClient calls.The multi-threaded test configuration with explicit TODO comment clearly documents why this change is needed—
CommitTask::new()uses blockingRpcClientcalls that require a multi-threaded runtime.programs/magicblock/src/magic_scheduled_base_intent.rs (1)
288-301: LGTM! Clean struct and conversion implementation.The
CommittedAccountstruct and itsFrom<CommittedAccountRef>implementation are well-structured and provide a clean conversion from the reference-wrapped internal representation to the owned public API type.test-integration/test-committor-service/tests/test_ix_commit_local.rs (2)
65-67: Multi-threaded runtime configuration applied consistently.The TODO comments clearly document the temporary workaround for blocking RpcClient calls in
CommitTask::new().
179-179: Strategy expectation updated toArgs.The test now expects
CommitStrategy::Argsinstead ofFromBuffer. This aligns with the new commit flow where smaller accounts can use args-based commits.test-integration/schedulecommit/test-scenarios/tests/01_commits.rs (1)
30-31: API updated to include deterministic seed parameter.The updated call to
get_context_with_delegated_committeesnow includes a seed parameter (b"magic_schedule_commit"), enabling deterministic test inputs. This aligns with the updated function signature in the test utilities.test-integration/test-committor-service/tests/utils/transactions.rs (1)
123-133: Pragmatic solution for dual commit paths.The special-case handling for "CommitState" to also match "CommitDiff" is a reasonable trade-off to avoid widespread test changes. The inline comment clearly explains the rationale. This allows existing tests that search for "CommitState" to pass regardless of which commit instruction type is actually used.
test-integration/test-committor-service/tests/common.rs (1)
152-165: Constructor pattern improves API consistency.Replacing the direct struct literal construction with
CommitTask::new()provides a cleaner, more maintainable API. The constructor encapsulates the logic for deciding whether to fetch base account data, making the call site simpler.magicblock-committor-service/src/tasks/task_strategist.rs (1)
267-280: Test updated to use constructor pattern.The test correctly uses
CommitTask::new()instead of struct literal construction, consistent with the new API pattern across the codebase.test-integration/programs/schedulecommit/src/utils/mod.rs (2)
53-53: LGTM: Type change aligns with Solana's system instruction API.Changing
sizefromusizetou64matches the signature ofsystem_instruction::allocate(line 90), improving type consistency.
75-79: LGTM: Debug logging and correct allocate call.The added logging helps diagnose lamport-related issues, and passing
sizedirectly tosystem_instruction::allocateis correct since it expectsu64.Also applies to: 90-90
test-integration/Cargo.toml (1)
40-40: LGTM: Dependency updates align with PR requirements.The delegation-program rev bump (to
e8d03936) and the addition ofrkyvandstatic_assertionssupport the newCommitDiffflow and zero-copy serialization requirements mentioned in the PR.Also applies to: 60-62, 74-74, 93-93
Cargo.toml (1)
111-114: LGTM: Workspace dependency updates are consistent.The delegation-program rev bump matches the test-integration workspace, and the explicit
magicblock-aperturepath entry is standard for workspace members.test-integration/programs/schedulecommit/Cargo.toml (1)
11-12: LGTM: New dependencies support zero-copy serialization.Adding
rkyvandstatic_assertionsenables the zero-copyOrderBookimplementation and compile-time size checks mentioned in the PR.magicblock-committor-service/src/tasks/task_builder.rs (1)
92-96: LGTM: Constructor encapsulates initialization logic.Using
CommitTask::new()instead of struct literals is good practice. The constructor (defined inmod.rs) now handles the logic for fetching base accounts when size exceeds the threshold.magicblock-committor-service/src/tasks/args_task.rs (1)
47-47: LGTM: Instruction generation delegated to CommitTask.Calling
value.create_commit_ix(validator)properly encapsulates the logic for choosing betweenCommitStateandCommitDiffinstruction generation withinCommitTask.test-integration/schedulecommit/test-scenarios/tests/utils/mod.rs (1)
16-35: LGTM: user_seed parameter enables per-context PDA derivation.Threading
user_seedthrough to the context constructors and consolidating initialization (lines 28-32) improves test setup consistency. The debug prints are helpful for test utilities.
GabrielePicco
left a comment
GabrielePicco
left a comment
There was a problem hiding this comment.
Thanks @snawaz. LGTM!
taco-paco
left a comment
taco-paco
left a comment
There was a problem hiding this comment.
Looks like generally this is going in the right direction, but wanted to leave some comments right away:
- We need to have 2 separate type
CommitTask&CommitWithDiffTask.CommitTaskcould be renamed into something to highlight that it isn't diff - Decision of whether to commit with diff or not shall be done in
TaskBuilder, it has all necessary knowledge to choose betweenCommitTask&CommitWithDiffTask. CommitTask & CommitWithDiffTask contain already complete information to be executed on chain. This will allow to get rid from clumsy Options and boolean flags in CommitTask - Task is representation of some instruction for some program, it can't know anything about Rpc or some network configs we use, so the logic of fetching account data for computing the diff shall be placed into TaskInfoFetcher
|
|
||
| use crate::{config::ChainConfig, ComputeBudgetConfig}; | ||
|
|
||
| let chain_config = |
There was a problem hiding this comment.
Task shouldn't be aware of the RpcClient, not be able to fetch anything from chain. Task is sort of representation of some chain instruction we want to execute.
The whole fetching of necessary info shall be moved into TaskInfoFetcher and assembly of Task shall be moved into TaskBuilder
| pub commit_id: u64, | ||
| pub allow_undelegation: bool, | ||
| pub committed_account: CommittedAccount, | ||
| base_account: Option<Account>, |
There was a problem hiding this comment.
CommitTask maybe ambiguos name now since when we were implementing this, we didn't think about another commit option via diffs. Since task is representation of instruction that is performed on chain we need to add new type of task for diffs. CommitWithDiffTask.
Please also rename CommitTask -> CommitDataTask or something like that to highlight that it commit of the whole account state and not the diff
There was a problem hiding this comment.
This will allow to remove clumsy Option<Account> that doesn't really make sense in the context of Commit diff task. If it is commit diif, then account against which we compute diff shall be present, if it is not present(for some weird reason, likely this should be an error) we shall choose CommitState instead.
100% agree. The logic for fetching base accounts inside the task isn’t a final solution ... it’s a temporary workaround. In the subsequent PRs, this logic will be moved out and put somewhere else... probably in With that, there might be some small refactoring too. I might make it a different task (like you said, So yes, that’s already in plan. It’ll just happen in the next set of PRs. Apart from that, is there anything you want me to change? Something trivial that can be done in this PR itself? |
|
@snawaz Noted, let's then make this PR ready for review and I will go over it again. The thing is that at the moment its not ready to be merged into master as its very raw. I'd recommend: add buffers, then modify We can review buffer PR & merge it separatly into this one, after that I will swiftly rereview this one and we will merge it |
This PR does the following things:
CommitDiff. So now when a commit is scheduled, it triggers eitherCommitStateorCommitDiffin the delegation-program, depending on the account size and theCOMMIT_STATE_SIZE_THRESHOLD.CommitDiffis effectively an optimization that reduces the amount of data transferred when committing small changes in large accounts, which improves performance and reduces costs.args, notbufferaccount.OrderBookaccount (10 KB). I initially wanted to try a few MB, but realized currently that’s not possible yet.CommitDiffwas invoked (in the delegation-program) with onlydiff len = 286whereasdata len = 10240(see first screenshot).diff len = 8(encoding the size of account on ER and the number of offset-pairs), and logs a warning (see second screenshot).Update dlp.so
Updated
test-integration/schedulecommit/elfs/dlp.soto use latest version of DLP withCommitDiffix.Changes in the existing tests.
Two kinds of changes:
Since it is
CommitTaskwhich decides the exact DLP instructionCommitStateorCommitDiff, and many tests which earlier used to get executed asBufferTask, will now be executed asArgsTaskbecause large accounts with small changes can be executed asArgsTaskas long as the transaction size stays within limit. So made the related changes in the tests.For some tests, I made the following change:
It is because now
CommitTask::new()uses blockingRpcClientto fetch account from the base chain, so the tests fail because they run in a single-threaded environment, giving this error:Using
multi_threadwith 2 threads makes the tests work again.Screenshots
Summary by CodeRabbit
New Features
API
Tests
Chores