fix: adjust state computation policy for Eth RPC methods#7116
Conversation
WalkthroughThis PR introduces an environment-controlled policy to gate state recomputation when executed-tipset data is missing from storage. A new ChangesRPC State Recomputation Policy
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Possibly related issues
Possibly related PRs
Suggested reviewers
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
✨ Simplify code
Warning Review ran into problems🔥 ProblemsStopped waiting for pipeline failures after 30000ms. One of your pipelines takes longer than our 30000ms fetch window to run, so review may not consider pipeline-failure results for inline comments if any failures occurred after the fetch window. Increase the timeout if you want to wait longer or run a Comment |
hanabi1224
force-pushed
the
hm/state-compute-rpc
branch
from
dda3dac to
42586fe
Compare
hanabi1224
requested review from
LesnyRumcajs and
akaladarshi
and removed request for
a team
LesnyRumcajs
left a comment
LesnyRumcajs
left a comment
There was a problem hiding this comment.
This is a functional breaking change, hence a CHANGELOG entry is required.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
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)
src/state_manager/state_computation.rs (1)
58-79:⚠️ Potential issue | 🟠 Major | 🏗️ Heavy liftKey the executed-tipset cache by recompute policy or provenance.
load_executed_tipset_for_rpcandload_executed_tipsetnow have different semantics, but this cache is still keyed only byts.key(). If any non-RPC path warms the cache through theAllowedbranch, a later RPC call with recomputation disabled will reuse that recomputed entry instead of returning the intended index-miss error. That makes the new policy depend on cache history rather than the configured setting.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/state_manager/state_computation.rs` around lines 58 - 79, The cache in load_executed_tipset_with_cache is keyed only by ts.key(), causing recompute policy to leak across callers (e.g., load_executed_tipset_for_rpc vs load_executed_tipset); fix by incorporating the recompute provenance into the cache key or by validating stored provenance before returning: modify the cache access sites (calls to self.cache.get(...) and self.cache.get_or_insert_async(...)) to use a composite key that includes ts.key() plus the Policy/ provenance (or keep a separate cache per provenance), or have ExecutedTipset carry provenance and check it against the current policy before returning and evict if mismatched. Ensure changes touch load_executed_tipset_with_cache, load_executed_tipset_for_rpc, and load_executed_tipset to preserve semantics.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/state_manager/state_computation.rs`:
- Around line 138-144: Current code unconditionally bails when event AMT is
unreadable unless allow_state_compute is true, causing RPCs that only need
state_root/message/receipt to fail; change the loader used in StateComputation
(the branch using allow_state_compute and the anyhow::bail! with
msg_ts.epoch()/msg_ts.key()) so it does not hard-fail for missing/unreadable
event trees when recomputation is disabled—instead return a lightweight tipset
state result that includes state_root/messages/receipts but leaves events as
None or a lazy-hydration handle; add a separate function (e.g.,
load_tipset_with_events or hydrate_events_for_tipset) that performs the
expensive AMT/event decoding and is only called by callers that need events
(update callers like Block::from_filecoin_tipset_with_full_tx and
eth_fee_history to call this explicit hydrator), and ensure allow_state_compute
still gates full hydration and the original bail remains inside that
full-hydration path.
---
Outside diff comments:
In `@src/state_manager/state_computation.rs`:
- Around line 58-79: The cache in load_executed_tipset_with_cache is keyed only
by ts.key(), causing recompute policy to leak across callers (e.g.,
load_executed_tipset_for_rpc vs load_executed_tipset); fix by incorporating the
recompute provenance into the cache key or by validating stored provenance
before returning: modify the cache access sites (calls to self.cache.get(...)
and self.cache.get_or_insert_async(...)) to use a composite key that includes
ts.key() plus the Policy/ provenance (or keep a separate cache per provenance),
or have ExecutedTipset carry provenance and check it against the current policy
before returning and evict if mismatched. Ensure changes touch
load_executed_tipset_with_cache, load_executed_tipset_for_rpc, and
load_executed_tipset to preserve semantics.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 84b3a48c-1360-4810-aca0-15c2aac7184f
📒 Files selected for processing (5)
docs/docs/users/reference/env_variables.mdscripts/tests/api_compare/docker-compose.ymlsrc/rpc/methods/eth.rssrc/rpc/methods/eth/filter/mod.rssrc/state_manager/state_computation.rs
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (1)
src/tool/subcommands/api_cmd/test_snapshot.rs (1)
68-68: ⚡ Quick winDocument the
unsafeblock and consider the LazyLock pattern.The
unsafe { std::env::set_var(...) }call lacks a comment explaining why it's acceptable. The existingINIT_RNG_SEEDpattern (lines 221–225) demonstrates best practice: use aLazyLockwith a comment noting thatstd::env::set_varis not thread-safe on Linux. Although the learning confirms that nextest runs tests in separate processes (mitigating cross-test races), the safety justification should still be documented for reviewers and maintainers.Additionally, consider adopting the
LazyLockpattern for consistency and to avoid redundantunsafecalls if this function is invoked multiple times.📝 Suggested improvement with LazyLock pattern
Add a static initializer following the existing pattern:
+ // Ensure FOREST_ETH_RPC_COMPUTE_STATE_ON_INDEX_MISS is set once per process. + // `std::env::set_var` is not thread-safe on Linux, so initialize once + // per process — never concurrently from parallel tests. + static INIT_ETH_RPC_COMPUTE_STATE: LazyLock<()> = LazyLock::new(|| { + if std::env::var("FOREST_ETH_RPC_COMPUTE_STATE_ON_INDEX_MISS").is_err() { + unsafe { std::env::set_var("FOREST_ETH_RPC_COMPUTE_STATE_ON_INDEX_MISS", "1") }; + } + }); + pub async fn run_test_from_snapshot(path: &Path) -> anyhow::Result<()> { - unsafe { std::env::set_var("FOREST_ETH_RPC_COMPUTE_STATE_ON_INDEX_MISS", "1") }; + LazyLock::force(&INIT_ETH_RPC_COMPUTE_STATE); let mut run = false;🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/tool/subcommands/api_cmd/test_snapshot.rs` at line 68, Document the unsafe std::env::set_var call and replace the ad-hoc unsafe call with a static LazyLock initializer similar to INIT_RNG_SEED: add a static (e.g., INIT_FOREST_ETH_RPC using std::sync::LazyLock) whose initializer performs the unsafe { std::env::set_var("FOREST_ETH_RPC_COMPUTE_STATE_ON_INDEX_MISS","1") } and include a comment explaining that setting env vars is not thread-safe on Linux and why a single-process/test-run justification still makes this safe; then remove the direct unsafe call and ensure callers trigger the static (consistent pattern with INIT_RNG_SEED).
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Nitpick comments:
In `@src/tool/subcommands/api_cmd/test_snapshot.rs`:
- Line 68: Document the unsafe std::env::set_var call and replace the ad-hoc
unsafe call with a static LazyLock initializer similar to INIT_RNG_SEED: add a
static (e.g., INIT_FOREST_ETH_RPC using std::sync::LazyLock) whose initializer
performs the unsafe {
std::env::set_var("FOREST_ETH_RPC_COMPUTE_STATE_ON_INDEX_MISS","1") } and
include a comment explaining that setting env vars is not thread-safe on Linux
and why a single-process/test-run justification still makes this safe; then
remove the direct unsafe call and ensure callers trigger the static (consistent
pattern with INIT_RNG_SEED).
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: fb980bcd-6620-47f4-b847-7ce191a03e1e
📒 Files selected for processing (3)
CHANGELOG.mdsrc/state_manager/state_computation.rssrc/tool/subcommands/api_cmd/test_snapshot.rs
✅ Files skipped from review due to trivial changes (1)
- CHANGELOG.md
🚧 Files skipped from review as they are similar to previous changes (1)
- src/state_manager/state_computation.rs
Codecov Report❌ Patch coverage is Additional details and impacted files
... and 3 files with indirect coverage changes Continue to review full report in Codecov by Sentry.
🚀 New features to boost your workflow:
|
|
Actionable comments posted: 0 |
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/state_manager/state_computation.rs`:
- Line 50: Update the tracking issue link in the inline comment that currently
points to "https://github.com/ChainSafe/forest/issues/7118" and change it to
"https://github.com/ChainSafe/forest/issues/7117" so the comment references the
correct follow-up issue for removing the test-only override (locate the comment
in state_computation.rs and replace the issue number).
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Repository UI
Review profile: CHILL
Plan: Pro
Run ID: 41f0f5c0-9c20-4237-928f-bc02dc9eeb5a
📒 Files selected for processing (1)
src/state_manager/state_computation.rs
2 participants
Summary of changes
Changes introduced in this pull request:
Reference issue to close (if applicable)
Closes
Other information and links
Change checklist
Outside contributions
Summary by CodeRabbit
Breaking Changes
New Features
FOREST_ETH_RPC_COMPUTE_STATE_ON_INDEX_MISSenvironment variable to enable state computation for Ethereum RPC methods when needed.Chores