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
4.3 KiB
Markdown
39 lines
4.3 KiB
Markdown
---
|
|
id: "bug-1583"
|
|
status: "fixed"
|
|
status_updated: "2026-09-27"
|
|
status_note: "70360ccf"
|
|
component: "apiserver"
|
|
severity: "unknown"
|
|
timestamp: "2026-08-18T13:35:00.000Z"
|
|
last_seen: "2026-09-19T09:00:00Z"
|
|
occurrences: 1
|
|
files: ["src/frontend/apiserver/backend/handler.go"]
|
|
tags: ["apiserver", "resval", "validation-gap", "data-integrity", "resource-value-verification", "handler.go", "OnResourceLocationUpdated"]
|
|
related: []
|
|
fix_commits: ["70360ccf"]
|
|
legacy: true
|
|
---
|
|
|
|
# bug-1583: Live task 8.1 run (resource-value-verification, comprehensive e2e validation): injecting resupdate…
|
|
|
|
## Description
|
|
Live task 8.1 run (resource-value-verification, comprehensive e2e validation): injecting resupdate -n totally-fake-resource-xyz -v 1 (a deliberately-unregistered resource name) alongside 9 genuinely-registered names -- the controller correctly rejects it (resval_typeresolve/update_resval's key_find(c->resmgr,...) returns -ENOENT, never written to w->host.resvals, confirmed empty via a direct resval query) and the scheduler correctly discards it (core_events.c's RESVALS_DELTA handler logs 'xce_core_push_events RESVALS_DELTA: unknown resource totally-fake-resource-xyz, skipping' at WARN and never applies it) -- but the apiserver's OnResourceLocationUpdated (src/frontend/apiserver/backend/handler.go:92) has NO such validation: it logged 'resource-location stored {"resource": "totally-fake-resource-xyz", "host": "resverify-worker", "value": 1}' and unconditionally called store.ResourceLocations.Put() for it, live-confirmed in apiserver.log.
|
|
|
|
## Environment
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## Reproduction
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## Root cause
|
|
do_update_resvals() (src/controller/opupd.c:10-51) fires XCE_CONTROLLER_HOOK_RESOURCE/HOOK_DETAIL_RESOURCE_UPDATE with the FULL, RAW, UNFILTERED xce_resval_list_t from the client RPC -- BEFORE update_resval_list() runs its own per-item validity check, and the HOOK_AFTER_EVENT fire (the one both sched_resource_hook_fun (sched.c:609-639) and apiserver_resource_hook_fun (apiserver.c:545-581) act on) still passes that same unfiltered list, since update_resval_list() only applies/logs per-item failures internally -- it never mutates or filters the list object itself. So both the scheduler and the apiserver receive a raw push containing EVERY submitted resource name, valid or not, regardless of whether the controller's own update_resval()->key_find(c->resmgr,...) accepted it. The scheduler's core_events.c independently re-validates each pushed name against its own resource table before applying (WARN + skip on unknown) -- but OnResourceLocationUpdated (handler.go:92-115) does zero such validation; it stores whatever ResourceName/Location/value it's handed, unconditionally. This is a genuine 3-way inconsistency: controller rejects+never-stores, scheduler rejects+discards+logs, apiserver silently accepts+stores forever.
|
|
|
|
## Fix
|
|
NOT FIXED -- discovered incidentally during resource-value-verification task 8.1's comprehensive live e2e validation run (deliberately injecting one unregistered resource name alongside the default 9-resource set to prove the run-level FAIL path). Explicitly out of task 8.1's boundary (validation-only, no code changes) and out of the resource-value-verification feature's boundary generally (design.md 'Out of Boundary' section explicitly excludes 'the controller's resval-forwarding hooks' and scopes handler.go's own change to exactly one logger.Debug() call after the pre-existing store.ResourceLocations.Put() -- the Put() call and its lack of validation both predate this feature). A real fix would add a registered-resource-name check to OnResourceLocationUpdated (or, more centrally, filter the notify() payload in do_update_resvals()/sched_resource_hook_fun/apiserver_resource_hook_fun to only include items update_resval_list() actually accepted) before persisting to store.ResourceLocations -- worth a dedicated follow-up task, does not block resource-value-verification's own completion (the skill's own verdict logic already correctly ignores/skip-checks rejected names via injection-results.tsv's accept/reject accounting, so this apiserver-side phantom-store gap does not affect task 8.1's or 7.1's PASS/FAIL correctness).
|
|
|
|
## Verification
|
|
_Not recorded (migrated from buglog.json)._
|
|
|
|
## References
|
|
_Not recorded (migrated from buglog.json)._
|