diff --git a/.kiro/specs/job-array-one-based-index/tasks.md b/.kiro/specs/job-array-one-based-index/tasks.md index 735b0122..cdcd34fa 100644 --- a/.kiro/specs/job-array-one-based-index/tasks.md +++ b/.kiro/specs/job-array-one-based-index/tasks.md @@ -252,3 +252,4 @@ - 8.1: client `format.ParseSelection(raw, allowRange)`, `ParseNameSelection(raw)`, `SplitSelections(list)`, `Selection.APIKey()`; query-range refusal rule tag "range" (`RuleQueryRange`); `ToAPIJobID(s, allowRange)`; peek uses `ToAPIJobIDLegacy`. For 8.2: `filters/job_filter.go` still runs `isValidJobId` before `ToAPIJobID` and both ID lists still split at every comma. - 10.1: meson 126/126 OK and every apiserver/xcecli Go package ok (asan, src/build.jai). No cgo unit test drives a full REST array submit (needs a controller); 10.2's live REST flow covers it. - 10.2: HA `--array-soak` PASS (1..N addressable after failover, one WAL record for 2 and 10000 elements); multihost PASS (`[1]` RUN→DONE on both hosts). Live REST via xce-deploy-ha + xcecli: `arr[1-3]` lists `[1]..[3]`; `bad[0-2]` rejected (rule 2.5); `kill [2]` and `kill -J 'arr3[3]'` reach only that element (a PEND element goes to EXIT alone). xcecli submit reads stdin when it is not a TTY: pass `1000-element arrays skipped rows or looped; fixed in efff6b65 (bug-1903 now partial: original GetJobsInfo symptom still open). Harness `setpend`/`jobrun` not-found gap logged as bug-1907 (controller acks first; outside this spec). Canonical gate: `xce-build.sh all` meson 126/126 + canonical Go packages ok; `apiserver/middleware` TestLoadJobs needs a live :8080 server with fixed credentials and is not in the canonical set. diff --git a/.wolf/buglog.json b/.wolf/buglog.json index 665fa013..7ac238a6 100644 --- a/.wolf/buglog.json +++ b/.wolf/buglog.json @@ -16839,9 +16839,9 @@ }, { "id": "bug-1903", - "status": "open", + "status": "partial", "status_updated": "2026-10-01", - "status_note": "newly logged", + "status_note": "page-token float64 precision loss for array element IDs fixed in efff6b65; the original GetJobsInfo repeat-first-page symptom with plain jobs is still unexplained", "component": "apiserver", "severity": "wrong-result", "timestamp": "2026-10-01T12:21:25Z", @@ -16849,8 +16849,8 @@ "occurrences": 1, "error_message": "apiserver GetJobsInfo page_token does not advance: the next page repeats the first\n\n**Observed:** found by the reviewer of task 7.3 (spec job-array-one-based-index) while comparing handler behaviour with HEAD in a unit-test harness: following the `page_token` returned by `GET /xce/v1/jobs` (`GetJobsInfo`) returns the same page again instead of the next one. Reproduced at HEAD (before task 7.3) with a plain filter `{submit_user:eq:all}` and a small `limit`, so it is not caused by the element-selection work.\n**Expected:** passing the returned `page_token` yields the next page of rows, and the last page returns no token.\n**Impact:** wrong result: clients that page through job listings (`xcecli jobs list` with paging, REST callers) see the first page repeatedly and never reach later jobs.\n**Found by:** review (task 7.3 of spec job-array-one-based-index, 2026-10-01)", "file": "src/frontend/apiserver/controllers/jobs.go, src/frontend/apiserver/store/query.go", - "root_cause": "TODO: required for fixed/partial", - "fix": "TODO: required for fixed/partial\n\nVerification:\n- Test: TODO: name + file; RED before the fix, GREEN after\n- Suites: TODO: suite results and build profile\n- Live: TODO: soak/deploy run, or n/a and why", + "root_cause": "Two separate parts:\n1. (found 2026-10-01, fixed) `store/query.go` `findPageTokenIndex` decoded the token with `json.Unmarshal` into `interface{}`, so `unique_job_id` became a float64. Array element IDs set bit 62, where float64 spacing is 1024, so the token no longer named the page's last row: paging skipped rows (N >= 1024) or found no match and restarted at row 0 (1001 <= N <= 1023). `queryControlRowsByJobID`, `FetchJobList` and `GetJobByIDs` page until the token is empty, so the restart case loops forever.\n2. (open) The originally observed repeat-the-first-page with plain jobs and a small `limit` in `GetJobsInfo` is not explained by part 1 for IDs below 2^53; cause not yet found.", + "fix": "Part 1: efff6b65. `findPageTokenIndex` decodes with `json.Decoder.UseNumber()` and converts integers to int64 (floats stay float64). Part 2: not yet fixed.\n\nVerification:\n- Test: (part 1) `TestPaginateRowsArrayElementIDsRoundTripPageToken` in `src/frontend/apiserver/store/query_test.go`; RED before the fix (2452 of 2500 rows), GREEN after\n- Suites: apiserver store/utils/controllers go test ok (CGO_ENABLED=0)\n- Live: n/a (unit-level; part 2 still needs a reproduction)", "tags": [ "apiserver", "query", @@ -16936,6 +16936,32 @@ "6e631caa" ], "record": ".wolf/bugs/bug-1906.md" + }, + { + "id": "bug-1907", + "status": "open", + "status_updated": "2026-10-01", + "status_note": "newly logged", + "component": "controller", + "severity": "wrong-result", + "timestamp": "2026-10-01T15:49:05Z", + "last_seen": "2026-10-01T15:49:05Z", + "occurrences": 1, + "error_message": "harness setpend and jobrun cannot report not-found: the controller acks status 0 first\n\n**Observed:**\nFound by code reading, no failing run captured. Both controller handlers reply before looking the job up:\n```\nsrc/controller/job.c:2246 handle_jobdispatch: init_reply_by_req(); reply.status = 0; xce_reply(...) /* then the work */\nsrc/controller/job.c:3869 do_set_preason: init_reply_by_req(); reply.status = 0; xce_reply(...) /* then the work */\n```\nSo harness `jobrun -j ` and `setpend -j ` succeed for a job (or array element) that does not exist.\n\n**Expected:** like the other single-job commands, `jobrun`/`setpend` on an unknown job ID report \"not found\" (job-array-one-based-index Requirements 4.7 and 5.13).\n**Impact:** wrong result: a test script that targets a wrong element (e.g. `$jobid[0]` on a new array) passes silently instead of failing.\n**Found by:** review (job-array-one-based-index task 4.3 implementation notes and feature validation, 2026-10-01)", + "file": "src/controller/src/job.c, src/harness/test.c", + "root_cause": "TODO: required for fixed/partial", + "fix": "TODO: required for fixed/partial\n\nVerification:\n- Test: TODO: name + file; RED before the fix, GREEN after\n- Suites: TODO: suite results and build profile\n- Live: TODO: soak/deploy run, or n/a and why", + "tags": [ + "harness", + "job-control", + "not-found", + "job-array" + ], + "related_bugs": [ + "bug-1896" + ], + "fix_commits": [], + "record": ".wolf/bugs/bug-1907.md" } ] } diff --git a/.wolf/bugs/bug-1903.md b/.wolf/bugs/bug-1903.md index d212e39b..019385a2 100644 --- a/.wolf/bugs/bug-1903.md +++ b/.wolf/bugs/bug-1903.md @@ -1,8 +1,8 @@ --- id: "bug-1903" -status: "open" +status: "partial" status_updated: "2026-10-01" -status_note: "newly logged" +status_note: "page-token float64 precision loss for array element IDs fixed in efff6b65; the original GetJobsInfo repeat-first-page symptom with plain jobs is still unexplained" component: "apiserver" severity: "wrong-result" timestamp: "2026-10-01T12:21:25Z" @@ -37,15 +37,17 @@ related: [] - Evidence: not kept (reviewer's temporary probe test, deleted) ## Root cause -TODO: required for fixed/partial +Two separate parts: +1. (found 2026-10-01, fixed) `store/query.go` `findPageTokenIndex` decoded the token with `json.Unmarshal` into `interface{}`, so `unique_job_id` became a float64. Array element IDs set bit 62, where float64 spacing is 1024, so the token no longer named the page's last row: paging skipped rows (N >= 1024) or found no match and restarted at row 0 (1001 <= N <= 1023). `queryControlRowsByJobID`, `FetchJobList` and `GetJobByIDs` page until the token is empty, so the restart case loops forever. +2. (open) The originally observed repeat-the-first-page with plain jobs and a small `limit` in `GetJobsInfo` is not explained by part 1 for IDs below 2^53; cause not yet found. ## Fix -TODO: required for fixed/partial +Part 1: efff6b65. `findPageTokenIndex` decodes with `json.Decoder.UseNumber()` and converts integers to int64 (floats stay float64). Part 2: not yet fixed. ## Verification -- Test: TODO: name + file; RED before the fix, GREEN after -- Suites: TODO: suite results and build profile -- Live: TODO: soak/deploy run, or n/a and why +- Test: (part 1) `TestPaginateRowsArrayElementIDsRoundTripPageToken` in `src/frontend/apiserver/store/query_test.go`; RED before the fix (2452 of 2500 rows), GREEN after +- Suites: apiserver store/utils/controllers go test ok (CGO_ENABLED=0) +- Live: n/a (unit-level; part 2 still needs a reproduction) ## References - Related: bug-1898 (found while fixing it) diff --git a/.wolf/bugs/bug-1907.md b/.wolf/bugs/bug-1907.md new file mode 100644 index 00000000..058db897 --- /dev/null +++ b/.wolf/bugs/bug-1907.md @@ -0,0 +1,58 @@ +--- +id: "bug-1907" +status: "open" +status_updated: "2026-10-01" +status_note: "newly logged" +component: "controller" +severity: "wrong-result" +timestamp: "2026-10-01T15:49:05Z" +last_seen: "2026-10-01T15:49:05Z" +occurrences: 1 +files: ["src/controller/src/job.c", "src/harness/test.c"] +tags: ["harness", "job-control", "not-found", "job-array"] +related: ["bug-1896"] +--- + +# bug-1907: harness setpend and jobrun cannot report not-found: the controller acks status 0 first + +## Description +**Observed:** +Found by code reading, no failing run captured. Both controller handlers reply before looking the job up: +``` +src/controller/job.c:2246 handle_jobdispatch: init_reply_by_req(); reply.status = 0; xce_reply(...) /* then the work */ +src/controller/job.c:3869 do_set_preason: init_reply_by_req(); reply.status = 0; xce_reply(...) /* then the work */ +``` +So harness `jobrun -j ` and `setpend -j ` succeed for a job (or array element) that does not exist. + +**Expected:** like the other single-job commands, `jobrun`/`setpend` on an unknown job ID report "not found" (job-array-one-based-index Requirements 4.7 and 5.13). +**Impact:** wrong result: a test script that targets a wrong element (e.g. `$jobid[0]` on a new array) passes silently instead of failing. +**Found by:** review (job-array-one-based-index task 4.3 implementation notes and feature validation, 2026-10-01) + +## Environment +- OS: no /etc/os-release; Docker container; NixOS kernel; kernel Linux 6.18.33 x86_64 +- libc: glibc 2.42-47 (interpreter /nix/store/wb6rhpznjfczwlwx23zmdrrw74bayxw4-glibc-2.42-47/lib/ld-linux-x86-64.so.2) +- Toolchain: gcc (GCC) 15.2.0, meson 1.9.1 +- Build: asan (src/build.jai) +- Code: scheduler-framework @ efff6b65-dirty +- Deployment: any (controller request handling) + +## Reproduction +1. Start a controller (xce-deploy-ha) and run `xce_test` against it. +2. `jobrun -j 0-0-1-5` (an ID that was never submitted) and `setpend` with `-j 0-0-1-5`: both report success (expected from the code; not run). +3. Compare `jobkill -j 0-0-1-5`, which reports not found. +- Frequency: deterministic (by code; not run live, 2026-10-01) +- Evidence: not kept (code references above) + +## Root cause +TODO: required for fixed/partial + +## Fix +TODO: required for fixed/partial + +## Verification +- Test: TODO: name + file; RED before the fix, GREEN after +- Suites: TODO: suite results and build profile +- Live: TODO: soak/deploy run, or n/a and why + +## References +- Related: none diff --git a/src/scheduler/src/repl.c b/src/scheduler/src/repl.c index d2e15cd1..c9db0972 100644 --- a/src/scheduler/src/repl.c +++ b/src/scheduler/src/repl.c @@ -139,11 +139,8 @@ int repl_split_tokens(const char* line, char tokens[][XCE_REPL_TOKEN_LEN], /* bug-1596 (task 12.5): strip one matching pair of surrounding * quote characters -- an operator pasting a value straight out of * another command's own output (e.g. a jobid from `info jobs`) and - * habitually quoting it otherwise gets a silent wrong-answer - * failure downstream (parse_jobid()'s whole-string-anchored regex - * rejects the literal quote characters, and a numeric strtoull() - * fallback then parses the leading quote as a nonnumeric prefix, - * silently producing 0) instead of an intelligible error. No + * habitually quoting it would otherwise have parse_jobid() reject + * the value because of the literal quote characters. No * escaping/embedded-space support -- just symmetric outer-quote * removal, sufficient for every argument this REPL's commands * accept (jobids, breakpoint filters, names). */