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>
4.3 KiB
| id | status | status_updated | status_note | component | severity | timestamp | last_seen | occurrences | files | tags | related | legacy | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| bug-1752 | fixed | 2026-09-24 | worker | unknown | 2026-09-01T10:00:00Z | 2026-09-19T09:00:00Z | 1 |
|
|
|
true |
bug-1752: worker-rejoin-resync task 6.1 (worker-side config-delivery gap): tasks.md's own task text says to…
Description
worker-rejoin-resync task 6.1 (worker-side config-delivery gap): tasks.md's own task text says to have 'the worker's existing config-receive handling call xcecd_finished_job_retention_set_window_ms() with the received value' -- premised on some existing per-field config-receive dispatch already existing at the worker that just needed one more field wired in. Investigating before implementing (reading src/worker/src/controller_connector.c's myreq_handler dispatch switch and src/lib/xcectx.c's vtask_entry, which decodes every inbound request before myreq_handler ever sees it) found this premise false: myreq_handler's switch had NO case for OPCode_push at all (only OPCode_startjob/runtasks/jobcontrol/fetchfile/jobStateQuery existed) -- a config push (Request_body_configpushmsg, the body_which xce_pushconfig_req()/worker_push_config() send) fell straight through to default: break; and was silently discarded in its entirety, for EVERY field including the pre-existing clustername, not just the not-yet-added workerResyncWindowMs. xcecd_finished_job_retention_set_window_ms()'s own doc comment (xcecd_worker_utils.h, task 4.1) had already independently found and documented the same gap ('the worker daemon has no runtime delivery path for this ClusterConfig field today... worker_push_config() only ever pushes hosttype/hostmodel/resource confvars via collect_confvars()... this setter is the seam a later task wires the real operator-configured value into once that delivery path exists') -- task 6.1 is that later task, but its own task text still assumed the delivery path itself already existed when it did not.
Environment
Not recorded (migrated from buglog.json).
Reproduction
Not recorded (migrated from buglog.json).
Root cause
No prior worker-rejoin-resync task (nor any task before this spec) ever needed the worker to consume ANY field off an OPCode_push config push, so myreq_handler's dispatch switch was never extended to handle that opcode at all -- xcectx.c's vtask_entry only special-cases the OTHER OPCode_push shape (Request_body_pushmsg, the generic job/worker/resource list-table push used for scheduler/apiserver connections) before handing off to the registered handler; a Request_body_configpushmsg-shaped OPCode_push passes through unmodified into myreq_handler's default: break; case. This predates worker-rejoin-resync entirely and was simply never exercised or noticed because nothing at the worker previously needed any pushed ClusterConfig field.
Fix
FIXED (2026-09-01, task 6.1, found live during implementation, same design-vs-reality-correction pattern as bug-1746/1747/1748/1751 earlier in this spec). Added a new case OPCode_push: to myreq_handler's dispatch switch, routing to a new handle_pushconfig() function that checks body_which == Request_body_configpushmsg and configdata->data_which == ConfigData_data_workerconfig before reading conf4worker->worker_resync_window_ms and calling xcecd_finished_job_retention_set_window_ms() with it -- scoped narrowly to closing the one seam task 6.1 needs (workerResyncWindowMs), not a general config-receive rework: clustername/confvars remain unconsumed by the worker, unchanged and explicitly out of this task's boundary. New tests in src/worker/testsuite/test_controller_connector.c (white-box, real wire encode via xce_pushconfig_req + real xce_decode + myreq_handler dispatch) cover: the real value reaching the retention setter exactly once, a second/different value also applied (proves live read, not memoized), a non-workerconfig ConfigData arm being ignored, and a NULL configpushmsg being a safe no-op.
Verification
Not recorded (migrated from buglog.json).
References
- Related: bug-1746, bug-1747, bug-1748, bug-1751