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>
3.7 KiB
| id | status | status_updated | status_note | component | severity | timestamp | last_seen | occurrences | files | tags | related | legacy | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| bug-1807 | fixed | 2026-09-24 | harness | unknown | 2026-09-14T15:50:00Z | 2026-09-19T09:00:00Z | 1 |
|
|
true |
bug-1807: harness_unit_test's test_test_cli.c: 4 of 132 tests deterministically FAIL every run…
Description
harness_unit_test's test_test_cli.c: 4 of 132 tests deterministically FAIL every run -- test_migrated_commands_sample_reject_unrecognized_flag (jobkill unrecognized-flag rejection), test_qctrl_hostctrl_confirm_gate_destructiveness_determination (qctrl -c or hostctrl -c prompt missing -- WHICH one shifts between builds), test_destructive_commands_prompt_by_default_and_yes_bypasses (upgradeconfig prompt missing), test_confirm_gate_session_toggle_skips_and_restores ('set +y' restore prompt missing). Each failing assertion burns its full 5000ms clitest_wait_for_text() timeout (total suite runtime ~28-29s vs ~9s once fixed) -- not a race, a genuine miss every time.
Environment
Not recorded (migrated from buglog.json).
Reproduction
Not recorded (migrated from buglog.json).
Root cause
CONFIRMED via isolated pty repro (bypassing Unity/gdb entirely): every cmd_() REPL command handler resets getopt_long() state with optind = 1; opterr = 0; before its own parse loop. glibc's getopt_long only re-initializes its internal static parser state (nextchar, first_nonopt/last_nonopt) when optind==0 OR on the process's very first-ever getopt_long() call (guarded by the internal _getopt_initialized flag) -- optind=1 alone is NOT sufficient once any earlier call in the same process has already run once. Since xce_test's REPL dispatches many independent cmd() calls (each with its OWN, differently-sized argv array) across the lifetime of one process, a command whose parse ends by consuming a short no-argument flag as its LAST token (e.g. qctrl -q normal -a, hostctrl -h node01 -o) leaves stale internal state that corrupts the NEXT unrelated command's getopt_long() call -- observed corrupting the NEXT command so badly that argv[0] (the command's own name, e.g. 'upgradeconfig' or 'jobkill') is misread as an unrecognized flag via flagspec_reject(). 100% reproducible in isolation: qctrl -q normal -a immediately followed by ANY other command (upgradeconfig, jobkill) corrupts that next command every time; the same session's commands run individually/first are unaffected. Explains why the specific victim test shifted between builds (bug-1807 was first seen failing on hostctrl -c, then qctrl -c after an unrelated same-file edit) -- runtime sequence-dependent, not random.
Fix
FIXED. Changed all 20 occurrences of optind = 1; in src/harness/test.c and the 1 occurrence in src/harness/join_worker_args.c to optind = 0; (the documented glibc idiom that forces getopt_long to fully reinitialize, including nextchar/first_nonopt/last_nonopt, on every independent parse). Verified: (1) the 3 isolated minimal pty repros that previously reproduced the corruption 100% of the time now pass cleanly; (2) harness_unit_test run 10x consecutively under the debug profile: 132 Tests 0 Failures every run (was 4 Failures every run pre-fix), runtime dropped from ~28-29s to ~8-9s per run, confirming the failures were full timeout burns, not near-miss races; (3) full /xce-build harness asan (meson test, ASAN-instrumented) also reports 5/5 OK including harness_unit_test, no sanitizer findings.
Verification
Not recorded (migrated from buglog.json).
References
Not recorded (migrated from buglog.json).