Repository navigation
Conversation
d0f9127 to
5c93ce3
Compare
|
The |
e8fec16 to
cf3473a
Compare
There was a problem hiding this comment.
Shipfox review
2 P2
The PR adds replicated capability-policy and proposal declaration fields plus pure validation helpers, then adds a pre-Raft declaration check. The admission barrier and replay floor are not fully enforced or preserved across normal policy updates and recovery.
Reviewed 34ab25b830cf3fb9c83ddc9662accd1b075b3248...cf3473a1ee663ce1ff6e2c5c914c4eb3902aaed0.
93d9f3d to
3282162
Compare
|
Fixed in commit 6d76261. The failing Tests-E2E-Business, Tests-E2E-Cluster, Tests-Model, and Tests-Antithesis-Workload jobs all exposed the same regression: Admission.Barrier serialized a semantic proposal without declaring its required capabilities, so the final proposal validator rejected it. Barrier now stamps the semantic capability field before marshaling and copies the derived declaration to the Raft proposal. Added a regression test that validates the final wire declaration. |
|
Fixed in commit e9899e8. The Dirty job was caused by seven protobuf artifacts generated with a local protoc/vtproto toolchain different from CI. These artifacts were regenerated with nix develop .#ci and the diff is limited to the CI-compatible generated output. |
|
Fixed in commit c3a8fd9. The Dirty job also reported lint --fix drift in eight Go files. The CI lint formatting/import changes are now committed; just pre-commit completes cleanly with 0 issues. |
c3a8fd9 to
635aee8
Compare
There was a problem hiding this comment.
Shipfox review
No new issues in 1 file changed since c3a8fd9.
8 earlier findings are still open.
Reviewed the changes since c3a8fd915a66caf89f1a4733c2f69fdb38b6af9a in f42c74fe6c770d907fe10bf6de7ecdc5c6220a2a...1e1fe85246fd7390b2b65fa40ea673fd44840320.
1faef2b to
c8ad6f7
Compare
There was a problem hiding this comment.
Shipfox review
New since 1e1fe85: 1 P1 · 1 P2
Resolved 1 earlier finding. 7 earlier findings are still open.
Reviewed the changes since 1e1fe85246fd7390b2b65fa40ea673fd44840320 in 3d73377506b392655c48f898cf87c93fec86cceb...c8ad6f7899ad9106d93cbdd266abbdda0182067c.
There was a problem hiding this comment.
Shipfox review
New since c8ad6f7: 1 P1
Resolved 1 earlier finding. 8 earlier findings are still open.
Reviewed the changes since c8ad6f7899ad9106d93cbdd266abbdda0182067c in 3d73377506b392655c48f898cf87c93fec86cceb...149dfff81c7adc894e1c73e2df38cba2af835bf3.
a99d84c to
c27bde1
Compare
flemzord
left a comment
There was a problem hiding this comment.
DECISION: APPROVE
Head reviewed: cb1af52
Base: 20c55fd
Incremental re-review from 3a2d277, including the rebased full PR context. The range-diff preserves the previously reviewed implementation; the functional delta fixes startup and background RPC deadline handling and narrows activation detection to the dedicated policy payload emitted by admission. The generated protobuf follow-up changes only its compiler-version header.
FIXED: the remaining bounded restart-refresh finding. Both local attempt expiry and an early remote DeadlineExceeded remain transient while the lifecycle is live. Repeated remote deadlines now retry through success; lifecycle cancellation and permanent rejection remain terminal. The background retry and comment-contract findings are also fixed. The earlier predicted-index capacity finding remains fixed. Native activation extraction, cancellation-safe membership fencing, recovered-peer refresh, and unsupported-capability recovery checks remain intact.
Validation passed in pinned Nix: agent-check; capability contract; revision-40 wire compatibility; protobuf lint; complete race suites for bootstrap, admission, capabilities, node, infrastructure/application membership, plan, state, policy processing, and reconciler. Refresh/error-classification regressions passed 30 race-enabled repetitions. Focused race checks passed for gRPC, backup policy, protobuf JSON, and validation gate selection. The evaluator benchmark and full-PR whitespace check passed. Tracked files are unchanged.
Limits: agent-check-pr gate selection was inspected with the exact base; its broad/full, E2E, and normalization gates were not executed in this read-only review. Released-binary upgrades, an activation-specific incremental-restore delta, and complete admission latency were not exercised locally. No .codegraph directory is present; source and caller inspection used Git and rg. Immediately before publication, the PR was open, non-draft, conflict-free, and unchanged at the reviewed head. Repository Invariants passed; Default and PR Images are still running.
Blocking findings: 0
Non-blocking findings: 0
Residual risk: MEDIUM
There was a problem hiding this comment.
Shipfox review
No new issues in 11 files changed since 266722e.
Resolved 1 earlier finding.
Reviewed the changes since 266722e8e42818bd25e44d31bb4b2b7c7c14f518 in 20c55fd8538630599b2dd03b3489454792263798...190b172fc13ae92aa249820913f8af5cf7728736.
Summary
rawNode.ProposeValidation
bash scripts/agent-checkgo test ./internal/infra/capabilities ./internal/infra/node ./internal/application/admissionRelease sequencing
RefreshMembershipcompatibility retry for older peers is intentionally retained.Relates to EN-3046. This PR is intentionally scoped to Ledger; the coordinated operator upgrade in EN-3047 is out of scope.