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>
4.8 KiB
| id | status | status_updated | status_note | component | severity | last_seen | occurrences | files | tags | related | legacy | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| bug-1828 | fixed | 2026-09-24 | scheduler | unknown | 2026-09-19T09:00:00Z | 1 |
|
|
true |
bug-1828: IMPACT CORRECTED 2026-09-18 (independent review of scheduler-sim-live-allocation task 2.3): this…
Description
IMPACT CORRECTED 2026-09-18 (independent review of scheduler-sim-live-allocation task 2.3): this is NOT just a display/accounting gap -- it meant host-scoped resource limits were NEVER ACTUALLY ENFORCED, in ANY mode (normal or simulation), for this function's entire history. Originally discovered while writing scheduler-sim-live-allocation spec task 2.3's own required TDD test (a host-scoped limit match, via the simulation-mode-registered limit plugin, must create a real usage instance verifiable through limit_get_usage_instances()). limit_prepare_allocation()/match_job_host_limit_() consistently returned rc=-2 (-ENOENT) instead of >= 0, and no usage instance was ever created, even though match_limit() itself genuinely matched (rc=1) and is_host_based_limit() correctly identified the limit as host-based. Tracing the call chain (plugin.c's plugin_prepare_allocation() hook loop -> sched_fill_limit_requests()) shows the missing usage instance meant a host-scoped limit's vres was never added to obj->hlimits, so sched_fill_limit_requests() always iterated an empty list -- the limit's numeric cap was silently never applied to any job's resource request, in production, since this code existed. This does not corrupt the dispatch/pend decision machinery itself (plugin_prepare_allocation()'s hook-calling loop discards the hook's return value, so the bug never fed into a control-flow branch) -- it is a dormant-limit-never-binds defect, but a real enforcement gap, not merely a cosmetic one.
Environment
Not recorded (migrated from buglog.json).
Reproduction
Not recorded (migrated from buglog.json).
Root cause
match_job_host_limit_() (src/scheduler/src/plugins/limit.c) is the host-scoped sibling of match_job_global_limit_() and has the same defect CLASS bug-1786 fixed in that function alone (misinterpreting key_find()'s negative "not found" return as fatal, rather than the correct not-yet-instantiated signal) -- not byte-for-byte identical wording (bug-1786's own guard was if (rc <= 0) { return rc; }; this sibling's was if (rc < 0) { return rc; }), same underlying mistake. Right after a successful match_limit(), it does rc = key_find(privdata->instances, vresname, &vres); if (rc < 0) { return rc; } before the if (vres == NULL) { instantiate_limit(...) } block below it. On the FIRST-EVER lookup of a given vresname (always true the first time a host-scoped limit ever matches), key_find() signals "not found" as a NEGATIVE rc (-ENOENT), not 0 -- this codebase's established kvmgr convention (get_group(), is_host_based_limit(), and bug-1786/bug-1495's own root causes) -- so this early-return unconditionally caught the expected miss and returned before instantiate_limit() could ever run. bug-1786 (2026-09-07) fixed this exact defect class in match_job_global_limit_() but never touched its host-scoped sibling, which retained the bug because no test exercised match_job_host_limit_()/limit_prepare_allocation() with real matching until this task added one (verified via grep: zero prior test call sites for either function anywhere in the test suite).
Fix
FIXED (2026-09-18, scheduler-sim-live-allocation task 2.3). Removed the erroneous if (rc < 0) { return rc; } guard in match_job_host_limit_() entirely, mirroring match_job_global_limit_()'s own bug-1786 fix exactly: vres (calloc'd NULL, only ever reassigned by a successful key_find) is already the correct, sufficient signal for "needs instantiation" via the existing if (vres == NULL) block immediately below. TDD: new test test_limit_prepare_allocation_matches_host_limit_via_sim_registered_plugin (test_repl_limit_table.c) drives the real limit_prepare_allocation()/match_job_host_limit_() path against a plugin registered via plugin_register_core() and a real percent-based host-scoped limit; RED (rc=-2, confirmed via temporary debug instrumentation tracing is_host_based_limit()=1, match_limit()=1, but limit_prepare_allocation()=-2) before this fix, GREEN (rc>=0, usage instance created and visible via limit_get_usage_instances()) after. Full scheduler suite: 79/79, zero regressions.
Verification
Not recorded (migrated from buglog.json).
References
Not recorded (migrated from buglog.json).