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>
3.5 KiB
| id | status | status_updated | status_note | component | severity | timestamp | files | tags | related | legacy | gitea_issue | gitea_synced_status | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| bug-1851 | fixed | 2026-09-24 | controller | crash | 2026-09-24T07:40:00Z |
|
|
true | 293 | fixed |
bug-1851: Live sequential HA soak (run-ha-soak.sh --sequential 6, round 5): newly-promoted primary ctrl-3…
Description
Live sequential HA soak (run-ha-soak.sh --sequential 6, round 5): newly-promoted primary ctrl-3 (term 10) died ~8s after promotion with ASAN heap-use-after-free READ in strdup <- collect_resval (dataex.c:71) <- apiserver_worker_change (apiserver.c:450) <- apiserver_worker_hook_fun <- xce_run_hooks <- persist_notify_worker_status_updated (persist.c:7997) <- resync_resolve_round (resync.c:418) <- resync_arm_round (resync.c:315) <- apply_join_state (worker.c:351) <- rejoin_worker (worker.c:444) <- handle_worker_join. Freed by the same thread in rejoin_worker (worker.c:434, xstring_dereference of the old hostmodel). Soak surfaced it as 'FAIL: round 5: no new primary elected after kill (no serving primary at term>9)' -- the new primary crashed rather than failing to elect.
Environment
Not recorded (migrated from buglog.json).
Reproduction
Not recorded (migrated from buglog.json).
Root cause
update_val_res() (worker.c:~99) memcpy()s the xce_val_t into w->host.resvals, so the 'type'/'model' string resources seeded by add_hostattrs_resvals() ALIAS the interned xstrings owned by w->host.attrs->hosttype/hostmodel (no reference taken). rejoin_worker() dereferences (frees, at refcount 0 -- even when the new value is the same string, since it dereferences before makeString) the old hosttype/hostmodel, then apply_join_state() calls resync_arm_round() BEFORE add_hostattrs_resvals() re-seeds resvals. When the resync round resolves synchronously inside arm (resync_resolve_round -> persist_notify_worker_status_updated -> worker hooks), apiserver_worker_change -> collect_resval strdup()s the dangling svalue. Latent since the alias was introduced; made reachable by worker-rejoin-resync's synchronous resolve+notify path on rejoin (bug-1841 area).
Fix
FIXED (2026-09-24, TDD). update_val_res() (worker.c) now stores a private strdup() for Value_u_svalue (copy taken before freeing the previous one, so val may alias v) and frees an old string when a slot changes type; the owning arrays now free their entries: w->host.resvals gets kill_resval (lib/wobj.c, replacing the FIXME NULL killer) and c->sharedinst's resval_obj_killer (cluster.c, previously a no-op) frees svalue + the xce_val_t -- both previously also leaked every xce_val_t. Also closes the same alias hazard for worker-reported string resvals (update_resval_list aliased the decoded request) and config values (mark_host_res). Regression test test_worker_join.c::test_rejoin_worker_string_resvals_survive_outstanding_round: first join arms a round (non-terminal job on host), second rejoin_worker with new attrs resolves it and fires a registered XCE_CONTROLLER_HOOK_WORKER hook that reads the type/model resvals -- RED reproduced the live stack exactly (heap-use-after-free, freed by rejoin_worker worker.c:432 xstring_dereference); GREEN after. Full meson test 109/109 OK. Note: get_resval() still hands out a borrowed svalue pointer (opget.c), now into the resval's own copy -- same lifetime contract as before, not widened.
Verification
Not recorded (migrated from buglog.json).
References
- Gitea: #293