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>
39 lines
3.3 KiB
Markdown
39 lines
3.3 KiB
Markdown
---
|
|
id: "bug-1626"
|
|
status: "fixed"
|
|
status_updated: "2026-09-27"
|
|
status_note: "0c1bac44"
|
|
component: "controller"
|
|
severity: "unknown"
|
|
timestamp: "2026-08-23T05:20:00Z"
|
|
last_seen: "2026-09-19T09:00:00Z"
|
|
occurrences: 1
|
|
files: ["src/controller/job.c"]
|
|
tags: ["controller", "concurrency", "locking", "rwlock", "RTTI_LOCK_WRITE", "job-state-mutation", "race-condition", "pre-existing"]
|
|
related: []
|
|
fix_commits: ["0c1bac44"]
|
|
legacy: true
|
|
---
|
|
|
|
# bug-1626: None yet observed at runtime -- found via code trace during /kiro-validate-design review of…
|
|
|
|
## Description
|
|
None yet observed at runtime -- found via code trace during /kiro-validate-design review of scheduler-dispatch-commit.
|
|
|
|
## Environment
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## Reproduction
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## Root cause
|
|
apply_job_fail_transition() (job.c:1804-1836) is explicitly documented as safe to call only "as long as the caller already holds job under RTTI_LOCK_WRITE" (a genuine rwlock, my_rwlock_wlock(), confirmed via src/lib/jobj.c:708-725's job_lock()). Its existing caller mark_job_fail() (job.c:1844) is itself called from job_ctrl() (job.c:965) at line 1083 -- but job_ctrl()'s only lock on the job object across its entire body (traced lines 965-1090) is lock(j, RTTI_LOCK_READ) at line 999, released via unlock(j) at line 1016, well before the mark_job_fail() call at line 1083. No RTTI_LOCK_WRITE is ever taken in this path. Separately, dispatch_a_job_() (job.c:831-963) mutates j->job.allocs under only RTTI_LOCK_READ on the containing maps (c->jobmgr, c->workermgr) and RTTI_LOCK_READ on the worker w -- never RTTI_LOCK_WRITE on the job j itself. RTTI_LOCK_READ is a genuine reader lock (my_rwlock_rlock()) that does not exclude concurrent readers from each independently mutating the same object, so if two threads process job-mutating requests for the same job concurrently, this is a real data race, not just an inaccurate comment. Contrast: many OTHER call sites in the same file (e.g. job.c:768, 1687, 1795, 2058, 2089, 2127, 2169, 2206, 2293, 2340, 2416) DO correctly take RTTI_LOCK_WRITE before mutating job state, confirming RTTI_LOCK_WRITE is this codebase's established, intentional convention for job-state mutation -- the job_ctrl()/mark_job_fail() and dispatch_a_job_() call sites appear to be inconsistent outliers, not a different intentional design.
|
|
|
|
## Fix
|
|
NOT YET FIXED -- discovered as a pre-existing gap while validating scheduler-dispatch-commit's design (that spec plans to add new mutations, including a WAL-fact-durable one, to dispatch_a_job_(), and reuse mark_job_fail() for its terminal dispatch-failure path -- both inherit this open question). Needs its own investigation: either (a) confirm some other serialization guarantee (e.g. single-threaded-per-job request processing) makes RTTI_LOCK_READ sufficient in practice and the RTTI_LOCK_WRITE comments/convention are overcautious, or (b) these call sites are missing a genuine lock upgrade and should take RTTI_LOCK_WRITE like their sibling call sites do. Flagged back to scheduler-dispatch-commit's design review as Critical Issue 1 rather than fixed inline, since it is a pre-existing condition this feature would otherwise silently inherit/extend, not something this feature caused.
|
|
|
|
## Verification
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## References
|
|
_Not recorded (migrated from buglog.json)._
|