Rongsong Shen
d26ff8870d
The buglog was one 27K-line JSON file whose fields covered only symptom,
root cause and fix, as free text: 2% of entries named the code version a
bug was found on, 12% the build profile, 36% a way to reproduce, and 58% of
the file was OpenWolf hook auto-detections ("Missing guard clause",
"Function not marked async" on .md files).
- .claude/skills/buglog/bug-record.md: the record standard. Eight
sections (Summary as the title, Description, Environment, Reproduction;
Root cause, Fix, Verification, References once fixed), each with the
content it must contain -- e.g. Environment needs the OS, the libc the
failing binary loads (from its ELF interpreter: glibc 2.42 in the nix
shell vs 2.40 on the host), toolchain, build profile and commit.
- .wolf/bugs/<id>.md: 500 records migrated from buglog.json; every field
verified to round-trip. They carry legacy: true and an explicit "not
recorded" where the old log had no such field; component derived from
file paths, severity from keywords (else unknown), fix_commits from
hashes in status_note/fix that git resolves (91 of 463 fixed).
- .wolf/buglog-archive.json: the 710 hook auto-detections, verbatim.
- .wolf/buglog.json is now an index generated from the records, keeping
the fields OpenWolf's hooks and `openwolf bug search` read.
- scripts/buglog.py rewritten: new (template + Environment collected from
the machine and --binary), check, status (refuses a status the record
doesn't support; fixed needs Root cause, Fix, Verification and
--commit hashes git resolves), note, recur, set, show, list, triage,
env, index. 29 unittest cases in scripts/tests/test_buglog.py.
- scripts/git-hooks/pre-commit runs `check --staged`: incomplete staged
records, a legacy record hand-changed to fixed without the fix-time
sections, a new record claiming legacy: true, or a stale index refuse
the commit.
- scripts/gitea-sync.py reads the records: issue body = the record's
sections plus linked fix commits (origin remote, GITEA_COMMIT_REPO to
override), and a body hash so an issue's text is refreshed when the
record changes.
- bug-1866 rewritten to the full standard as the worked example.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
4.6 KiB
| id | status | status_updated | status_note | component | severity | timestamp | last_seen | occurrences | files | tags | related | legacy | ||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| bug-1753 | fixed | 2026-09-24 | controller | unknown | 2026-09-01T20:30:00Z | 2026-09-19T09:00:00Z | 1 |
|
|
|
true |
bug-1753: worker-rejoin-resync task 6.2 (controller-side OPCode_jobStateReport dispatch): design.md's Error…
Description
worker-rejoin-resync task 6.2 (controller-side OPCode_jobStateReport dispatch): design.md's Error Handling section only anticipated the STANDBY-role case for this opcode pair ("verify... whether OPCode_jobStateQuery/OPCode_jobStateReport need an explicit redirect-reply path... Expected answer: no... but this must be confirmed, not assumed, given bug-1700's documented history of this exact class of gap recurring across two other opcodes"). Auditing ha_dispatch_gate() (ha.c) in full per that instruction found the STANDBY case was indeed already correctly handled automatically (OPCode_jobStateReport, left absent from the CTRL_STANDBY allow-list, falls to the generic default: return -XCE_ERR_STANDBY branch, and cluster.c's existing bug-1700 post-gate block already fills the redirect hint for any such rejection with zero opcode-specific code needed) -- but the same audit surfaced a DIFFERENT, unanticipated gap in the SAME function: a CTRL_PRIMARY controller that has temporarily lost write quorum rejects OPCode_jobStateReport outright with -XCE_ERR_NOT_READY, because the opcode was absent from ha_dispatch_gate's quorum-loss "fact-bearing ops" allowlist (OPCode_notify/OPCode_update/OPCode_setpreason/OPCode_clrpreason, ha.c ~line 4686-4696) that already exists specifically to prevent this exact class of loss (bug-396) for other worker/scheduler fact reports.
Environment
Not recorded (migrated from buglog.json).
Reproduction
Not recorded (migrated from buglog.json).
Root cause
OPCode_jobStateReport is sent fire-and-forget (xce_job_state_report_req(), request.c/xce_request()) with no redelivery on rejection -- structurally identical to OPCode_notify, the opcode bug-396's fix (ha-commit-barrier-rollout) originally added this allowlist to protect. Since worker-rejoin-resync's two new opcodes were added to controller.capnp's OPCode enum without also being added to ha_dispatch_gate's quorum-loss allowlist, a job-state reconciliation report arriving at a primary during a transient write-quorum-loss window (e.g. mid-failover, the exact scenario this whole feature exists to recover from) would be rejected outright and never retried by the worker -- reproducing bug-396's own loss mode for precisely the facts worker-rejoin-resync (bug-1741) exists to recover.
Fix
FIXED (2026-09-01, task 6.2, found live during the task's own mandated standby-redirect investigation -- not assumed safe, confirmed by reading the full gate function as instructed). Added case OPCode_jobStateReport: alongside the existing OPCode_notify/OPCode_update/OPCode_setpreason/OPCode_clrpreason cases in ha_dispatch_gate's quorum-loss fact-bearing-ops allowlist, so a report arriving while the primary temporarily lacks write quorum is admitted (not rejected) and can flow into the same RETAIN/pending-facts WAL-append mechanism those sibling opcodes already rely on. OPCode_jobStateQuery needed no corresponding change: the controller only ever SENDS that opcode (controller -> worker), never receives it, so it never reaches ha_dispatch_gate at all. New regression test test_gate_no_quorum_admits_jobstatereport_fact (test_commit_barrier_rollout.c), mirroring test_gate_no_quorum_admits_notify_fact's exact fixture/assertion shape; also added to test_gate_with_quorum_admits_all_opcodes's opcode list for parity. The CTRL_STANDBY case needed and received no code change at all -- proven correct as-is by three new tests in the new src/controller/testsuite/test_jobstatereport_dispatch.c (rejects-when-not-primary, standby-with-known-primary-redirects, standby-no-known-primary-no-redirect), mirroring test_sched_admin_dispatch.c's identical bug-1700 regression tests for OPCode_schedAdmin, the 'real existing primary-only opcode' ground truth this task's brief called for.
Verification
Not recorded (migrated from buglog.json).
References
- Related: bug-396, bug-1700, bug-1746, bug-1747, bug-1748, bug-1751, bug-1752