Adopt SQLite project unit test harness#692
Conversation
brandonpayton
force-pushed
the
integration/kad-wtb-sqlite-testing
branch
from
93ef2b4 to
92cd505
Compare
Phase B-1 matrix build status —
|
| Package | Arch | Status | Sha |
|---|---|---|---|
| libcurl | wasm32 | built | ce011fa0 |
| libcxx | wasm32 | built | cb86af2e |
| libcxx | wasm64 | built | 6a6ad11d |
| libpng | wasm32 | built | 72185039 |
| libxml2 | wasm32 | built | df935363 |
| libxml2 | wasm64 | built | 98da11bb |
| ncurses | wasm32 | built | 7e3c0e90 |
| openssl | wasm32 | built | a500b5dc |
| openssl | wasm64 | built | e4db922d |
| sqlite | wasm32 | built | 621a54bc |
| sqlite | wasm64 | built | 5bcb5605 |
| zlib | wasm32 | built | 40932d64 |
| zlib | wasm64 | built | 4ebfa8c2 |
| bash | wasm32 | built | 2f184714 |
| bc | wasm32 | built | 679b5b57 |
| bzip2 | wasm32 | built | d25eab2e |
| coreutils | wasm32 | built | 2f577527 |
| curl | wasm32 | built | 8958e33c |
| dash | wasm32 | built | 5ae935a3 |
| diffutils | wasm32 | built | 6283becf |
| dinit | wasm32 | built | c007ac0a |
| fbdoom | wasm32 | built | a07b454d |
| file | wasm32 | built | 02b42883 |
| findutils | wasm32 | built | 2264e878 |
| gawk | wasm32 | built | 263d59ae |
| git | wasm32 | built | 0398f88b |
| grep | wasm32 | built | 06cfacd3 |
| gzip | wasm32 | built | f1ea005c |
| kandelo-sdk | wasm32 | built | be92ea52 |
| kernel | wasm32 | built | 98e21a05 |
| less | wasm32 | built | a7db2115 |
| lsof | wasm32 | built | 23cde9ae |
| m4 | wasm32 | built | a8e81cc0 |
| make | wasm32 | built | 3ade5be8 |
| mariadb | wasm32 | built | b98ae7c6 |
| mariadb | wasm64 | built | ed07a0b9 |
| msmtpd | wasm32 | built | 263d33ef |
| nano | wasm32 | built | 844e9bbe |
| netcat | wasm32 | built | 72fd3e91 |
| nethack | wasm32 | built | 93ed633b |
| nginx | wasm32 | built | 8ef1852a |
| php | wasm32 | built | 31f3844d |
| posix-utils-lite | wasm32 | built | fa6d1074 |
| sed | wasm32 | built | 498dd764 |
| spidermonkey | wasm32 | built | fc097aa8 |
| tar | wasm32 | built | 20922362 |
| tcl | wasm32 | built | cb2a2699 |
| unzip | wasm32 | built | a27a6076 |
| userspace | wasm32 | built | 9eb2c94c |
| vim | wasm32 | built | b8621f5b |
| wget | wasm32 | built | d8367779 |
| xz | wasm32 | built | 8f63492f |
| zip | wasm32 | built | 0e37cb2a |
| zstd | wasm32 | built | ae08e187 |
| lamp | wasm32 | built | 4abe20e8 |
| mariadb-test | wasm32 | built | 3a1dce84 |
| mariadb-vfs | wasm32 | built | e655616c |
| mariadb-vfs | wasm64 | built | 255bfde1 |
| nethack-browser-bundle | wasm32 | built | b33d5459 |
| node | wasm32 | built | 288ce1f6 |
| rootfs | wasm32 | built | 04a457ae |
| spidermonkey-node | wasm32 | built | 2ac80e33 |
| vim-browser-bundle | wasm32 | built | 6c746258 |
| wordpress | wasm32 | built | 7255dced |
| shell | wasm32 | built | 2451b31b |
| node-vfs | wasm32 | built | f799a4df |
Auto-generated; replaced on each push. Raw data in the publish-status workflow artifact.
…d-wtb.20@mqe5fix' into HEAD
Return the active thread id for gettid/set_tid_address and keep clone helper state consistent in host tests. Add a wasm32posix pthread_create overlay that preserves 16-byte shadow-stack alignment after reserving startup arguments, so variadic calls in worker threads read 64-bit arguments correctly. Cover the SQLite sorter-temp failure mode with a pthread regression that exercises va_arg(uint64_t) and snprintf(%llx%c) from a worker thread.
SQLite project-unit harness interim status, 2026-06-18Tracking: This work is still active, but the PR had not been updated while the run was blocked by Kandelo platform crashes. This comment records the current state; it is not a pass claim and does not replace the final PR description update. Current scope
Platform blockers hit before the current retryThese were incomplete Kandelo platform crashes, not SQLite pass/fail results:
The platform fixes above still need to be reviewed as commits/PR updates before this PR can be considered publication-complete. Current live runCommand policy: timeout 86400s scripts/dev-shell.sh bash -lc 'scripts/run-sqlite-project-unit-tests.sh --host node --permutation all --jobs 32 --timeout-ms 86400000 --results-root test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-after-pathfix-20260618-1810 --workdir /Users/brandon/src/kandelo/test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-after-pathfix-20260618-1810/workdir --keep-workdir'
Next update expectedThe next PR update should happen when one of these occurs:
|
|
SQLite PR #692 audit update: Node Scope and policy:
Run outcome:
Stopped copied DB counts, diagnostic only:
Diagnostic failed rows, not final suite failures: Diagnostic running rows at stop: Artifacts:
Publication status:
|
|
SQLite PR #692 audit update: local mmap metadata-allocation fix is applied and the required fresh Node/browser Local fix summary:
Validation run locally:
Fresh
Explain counts:
Publication status:
|
kd-nbh SQLite PR #692 audit update: Node all/jobs32 crashed/wedged after mmap fixThis is an incomplete platform-blocker run, not a SQLite pass/fail result. The run passed the previous mmap trap point, then hit kernel/runtime traps and wedged with the Node process alive at 0%% CPU. I preserved artifacts, sent SIGINT, and the wrapper exited status 130. Command/policy:
Artifacts:
Stopped-state DB counts from copied DB (integrity_check=ok): State counts: Crash signature summary: first recorded trap was syscall=1 RuntimeError: unreachable at normalized command log line 176; repeated handleFcntlLock RuntimeError lines follow; the final cascade includes memory access out of bounds in syscalls 47/40/64/124. This is still a kernel/runtime blocker, not a SQLite failure classification. Decision: keeping kd-nbh open and investigating/fixing this blocker in-place before another full retry; no separate bead yet because it directly blocks the assigned audit. Complete stopped-state failed jobs (diagnostic only)Complete stopped-state running jobs at stop time |
kd-nbh SQLite PR #692 audit update: open-path kernel refresh and fresh all --explainThis is a milestone update, not a final SQLite pass/fail report. After the post-mmapfix run crashed first in Validation/build commands:
|
kd-nbh SQLite PR #692 audit update: Node all/jobs32 crashed after open-path fixThis is an incomplete Kandelo platform-blocker run, not a SQLite pass/fail result. The run passed the prior SYS_OPEN crash point, then hit a new kernel trap in SYS_GETDENTS64 and cascaded through fcntl/memory traps. I preserved artifacts, stopped the runner, copied the stopped DB/WAL/SHM, and queried only the copied database. Command/policy:
Artifacts:
Copied DB integrity: ok. Stopped-state diagnostic counts from copied DB: State counts: Crash signature summary:
The rows below are diagnostic stopped-state rows only because the platform crashed before suite completion. Complete stopped-state failed jobsComplete stopped-state running jobs at stop timeDecision: keeping kd-nbh open and investigating/fixing this blocker in-place before another full retry. If the root cause splits beyond the assigned audit, I will create/route a focused blocker bead and keep this PR updated. |
SQLite PR #692 audit: post-getdents kernel refresh + fresh all --explainThis is a milestone update for Root cause addressed since the previous crash/quarantine: the first post-open-pathfix platform trap was Validation before this explain:
Required fresh explain after source/kernel refresh:
Explain counts:
Scope/skips remain unchanged: selected public SQLite Next action: retry the selected public Tcl |
SQLite PR #692 audit: Node all/jobs32 after getdents fix hit new platform blockerThis run is incomplete due to a Kandelo platform crash, not a SQLite pass/fail result. Command/policy:
Artifacts:
Stopped-state diagnostic counts from copied DB: State counts: Crash signature summary: the run progressed to about Scope/skips remain unchanged: selected public SQLite Complete stopped-state failed rows, diagnostic onlyComplete stopped-state running rows at stop timeNext action: diagnose and fix the |
kd-nbh SQLite PR #692 audit update: fcntl-lock crash fix + fresh explainPublished milestone for convoy Root cause addressed since the last quarantine: the first post-getdentsfix trap was through Validation after source/kernel refresh:
Required fresh
Explain counts:
This explain run is enumeration only, not a pass/fail result. Scope/skips unchanged: selected public SQLite Tcl Next action: retry selected public Tcl |
kd-nbh SQLite PR #692 audit update: post-fcntl-lockfix Node run quarantinedThis is an incomplete Kandelo platform-blocker run, not a SQLite pass/fail result. The run passed the previous fcntl-lock crash window, then hit a new kernel/runtime trap cascade and was stopped for quarantine. Command/policy:
Artifacts:
Copied DB integrity: Stopped-state diagnostic counts only: State counts: Crash signature summary: latest visible progress before the first trap was This stopped-state failure list is diagnostic only because the platform crashed before the suite completed. Complete stopped-state failed jobs (diagnostic only)Complete stopped-state running jobs at stop timeSkips/out-of-scope unchanged: selected public SQLite Tcl Next action: diagnose the first trap from this quarantine, fix the platform root cause, rebuild the kernel if needed, rerun fresh Node/browser |
kd-nbh SQLite PR #692 audit update: close-path crash fix + fresh explainPublished milestone for convoy Root cause addressed since the last quarantine: the first post-fcntl-lockfix trap was Validation after source/kernel refresh:
Required fresh
Explain counts:
This explain run is enumeration only, not a pass/fail result. Scope/skips unchanged: selected public SQLite Tcl Next action: retry selected public Tcl |
kd-nbh SQLite PR #692 audit update: post-closefix Node run quarantinedThis is an incomplete Kandelo platform-blocker run, not a SQLite pass/fail result. The run passed the previous fcntl-lock crash window and reached the neighborhood of the previous close-path trap, then hit a new write-path kernel trap cascade and was stopped for quarantine. Command/policy:
Artifacts:
Copied DB integrity: Stopped-state diagnostic counts only: State counts: Crash signature summary: latest visible progress before the first trap was This stopped-state failure list is diagnostic only because the platform crashed before the suite completed. Complete stopped-state failed jobs (diagnostic only)Complete stopped-state running jobs at stop timeSkips/out-of-scope unchanged: selected public SQLite Tcl Next action: diagnose the first write-path trap from this quarantine, fix the platform root cause, rebuild the kernel if needed, rerun fresh Node/browser |
|
kd-nbh update after pipe-write fix/explain Milestone: fixed the next platform crash from the post-closefix Node run and completed the required fresh Root cause: the first trap in Change: Validation commands and counts:
Artifacts:
Skipped/out of scope remains unchanged: TH3/private SQLite suites and non-public SQLite test assets are out of scope for this public package audit. The selected public Tcl suite remains Next: retry Node public |
|
kd-nbh quarantine after post-pipewritefix Node all/jobs32 retry Milestone: new platform blocker/quarantine. The run is incomplete and must not be counted as a SQLite pass/fail result. Run command/policy:
First trap/signature:
Quarantine artifacts:
Stopped copied DB status:
Partial failed/running rows are captured in Next: diagnose and fix the exec-state/kernel crash before another full retry. Fresh Node/browser |
|
kd-nbh SQLite PR #692 audit update: post-exec/pipefix Node all/jobs32 stopped during testrunner control DB initialization This run is incomplete and is not a SQLite pass/fail result. It failed before any jobs launched because SQLite's testrunner control database became malformed while building the Command/policy:
Observed failure:
Counts:
Artifacts:
Stopped-copy diagnostics:
Next action: diagnose the testrunner-control-DB corruption path as the current platform blocker. This is separate from the earlier kernel trap signatures and still blocks completing the public Tcl |
|
2026-06-18 21:10 EDT diagnostic update for the post-exec/pipefix control-DB blocker. After the malformed
Interpretation: the malformed Skip/out-of-scope remains unchanged: TH3/private and other non-public SQLite suites are out of scope; selected public Tcl suite is |
|
2026-06-18 21:20 EDT additional kernel refresh before full retry: pipe-read allocation/wakeup path. Before restarting the full Node run, I found the current kernel artifact had advanced again after the previous diagnostics. The change is kernel-only: Validation for this source/kernel state:
Sequencing note: the full retry announced in the previous comment was stopped before producing audit evidence after this newer refresh became visible. That abandoned start is Next run policy remains unchanged: selected public SQLite Tcl suite |
|
2026-06-18 21:27 EDT run-state update after pipe-read refresh. One full Node retry was interrupted before completion and is not a SQLite pass/fail result:
There is now one active full retry, which I am treating as the current Node full run:
Skipped/out of scope unchanged: TH3/private and non-public SQLite suites remain out of scope; selected public suite remains |
kd-nbh SQLite PR #692 audit update: exec/CLOEXEC refresh + active retryPublication status: posting this milestone to PR #692 and mirroring it to bead Previous full retry quarantined as platform crash/incomplete:
Root cause addressed in the latest kernel refresh:
Validation after refresh:
Active retry:
Skipped/out-of-scope unchanged: TH3/private SQLite suites and non-public assets are out of scope for this Kandelo package audit. Selected public suite remains upstream |
kd-nbh SQLite PR #692 audit update: Node retry quarantined on Wasm memory-growth hangPublication status: posting this milestone to PR #692 and mirroring it to bead Run/quarantine:
Hang evidence:
Stopped diagnostic counts only:
Key artifacts:
Skipped/out-of-scope unchanged: TH3/private SQLite suites and non-public assets are out of scope. Selected public suite remains upstream Next action: diagnose the focused Wasm memory-growth hang path before retrying the full Node |
|
kd-nbh active retry update, 2026-06-18 22:06 EDT\n\nA Node public SQLite testrunner retry is currently running after the exec/CLOEXEC refresh and prior Wasm-memory-growth quarantine. This start was not yet published, so recording it now.\n\nCommand/policy:\n- Command: timeout 86400s scripts/dev-shell.sh bash -c "scripts/run-sqlite-project-unit-tests.sh --host node --permutation all --jobs 32 --timeout-ms 86400000 --results-root 'test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-cleanpatch-20260618-220351'"\n- Host/permutation/jobs: node / all / 32\n- Runner timeout: 86400000 ms\n- Outer hard timeout: 86400s\n- Timeout/crash/wedge policy: incomplete, not pass/fail\n- Public suite scope remains upstream SQLite testrunner.tcl all, 10523 discovered jobs from the last required Node/browser all --explain after the source/kernel refresh. TH3/private/non-public suites remain out of scope.\n\nArtifacts/current state:\n- Run root: test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-cleanpatch-20260618-220351\n- Live command log: test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-cleanpatch-20260618-220351/command.log\n- Live temp workdir from runner log: /tmp/kandelo-sqlite-official.vL0qZf\n- Caveat: this start did not pass explicit --workdir/--keep-workdir. I have identified the temp workdir and will preserve it before any stop/quarantine; I am not reading the live testrunner.db.\n- Current log-only progress at publication: at least 81/10523 jobs, f24, r32, with no RuntimeError/kernel threw/memory access out of bounds/unreachable/handleFcntlLock/malformed DB signatures in command.log. These visible FAILED lines are diagnostic only until the suite completes.\n\nNext milestone will be either normal runner completion with exact counts/failure catalog, or a new quarantine/blocker comment with preserved artifacts and stopped-state counts. |
|
kd-nbh progress milestone, 2026-06-18 22:12 EDT\n\nThe active Node public SQLite all/jobs32 retry has passed the previous exec/CLOEXEC-refresh quarantine point. Prior run test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-after-execcloexecfix-20260618-213916 was stopped around 186/10523 with CPU-active samples in V8 Wasm memory growth. Current run test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-cleanpatch-20260618-220351 has reached log-only progress 205/10523, f35, r32.\n\nStill not a pass/fail result:\n- Command/log root: test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-cleanpatch-20260618-220351\n- Temp workdir: /tmp/kandelo-sqlite-official.vL0qZf\n- Policy: runner_timeout_ms=86400000, outer timeout=86400s; timeout/crash/wedge remains incomplete, not pass/fail\n- I am not reading the live testrunner.db.\n- No RuntimeError/kernel threw/memory access out of bounds/unreachable/handleFcntlLock/malformed DB signatures are present in command.log at this checkpoint.\n\nCurrent interpretation: the earlier Wasm-memory-growth sample is not yet proven to be a hard platform wedge; this retry is being allowed to continue under the documented timeout policy until normal completion or a clear crash/wedge. |
|
kd-nbh interruption update, 2026-06-18 22:15 EDT The active Node retry was interrupted before completion and before the wrapper produced Interrupted run:
Next action: restarting the selected public Tcl |
|
kd-nbh restart confirmation, 2026-06-18 22:17 EDT Replacement Node public SQLite Command: Artifacts:
Policy unchanged: selected public SQLite Tcl |
|
kd-nbh platform blocker/quarantine, 2026-06-18 22:40 EDT The preserved-workdir Node public SQLite Command/policy:
First/current blocker signature:
Artifacts:
Stopped copied DB status:
The failed/running rows are diagnostic only because the parent control DB lock failure stopped the audit before suite completion. Next action: diagnose the testrunner control-DB locking path as the current Kandelo platform blocker before another full retry. Browser remains unvalidated for full execution; the last fresh Node/browser |
kd-nbh SQLite PR #692 audit update: control-DB lock cleanup fix + fresh explainLatest incomplete blocker addressed locally: the preserved-workdir Node Root-cause fix in this refresh:
Validation run after the fix:
Runtime artifacts refreshed:
Required fresh post-refresh explain:
Next action: retry selected public SQLite Tcl |
|
SQLite audit Status: still running, incomplete, not a pass/fail result. Command under audit: timeout 86400s scripts/dev-shell.sh bash -c "scripts/run-sqlite-project-unit-tests.sh --host node --permutation all --jobs 32 --timeout-ms 86400000 --results-root 'test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-after-lockexitfix-20260618-225026' --workdir '/Users/brandon/src/kandelo/test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-after-lockexitfix-20260618-225026/workdir' --keep-workdir"Timeout policy: runner timeout Artifacts:
Log-only checkpoint, without inspecting the live Signature scan of Process checkpoint: Node testfixture process |
|
SQLite audit Status: incomplete / platform blocker, not a SQLite pass/fail result. I stopped the run after it wedged: Command under audit: timeout 86400s scripts/dev-shell.sh bash -c "scripts/run-sqlite-project-unit-tests.sh --host node --permutation all --jobs 32 --timeout-ms 86400000 --results-root 'test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-after-lockexitfix-20260618-225026' --workdir '/Users/brandon/src/kandelo/test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-after-lockexitfix-20260618-225026/workdir' --keep-workdir"Timeout policy: runner timeout Artifacts:
Copied-DB integrity: Stopped-state job counts:
Stopped-state case counts from completed/failed job rows:
Complete failed-job list observed before the wedge: The 32 jobs still marked Signature scan of Skips/out-of-scope remain unchanged: TH3/private SQLite tests are skipped because they are not public package tests for this audit. Public |
kd-nbh SQLite audit update: raw wakeup host fix + fresh all explainThis milestone addresses the latest incomplete Node Root-cause fix in shared host runtime:
Validation:
Fresh required SQLite enumeration after the shared host source refresh:
Next action: start a new preserved Node public SQLite Tcl |
kd-nbh SQLite audit update: Node all/jobs32 retry active after raw wakeup fixA new preserved Node public SQLite Tcl Command and timeout policy:
Artifacts:
Live log-only checkpoint, no live DB inspection:
The selected public upstream scope remains |
|
SQLite PR #692 audit update for Command/timeouts:
What happened:
Stopped DB copy counts (
Artifacts:
Publication status: PR #692 updated here; bead notes update follows. Next step is to isolate/fix the runner DB lock/platform locking blocker before another full |
|
SQLite PR #692 audit update for Change made:
Why this targets the blocker:
Validation run after the fix:
Publication status: PR #692 updated here; bead notes update follows. Next step: retry Node |
|
SQLite PR #692 audit update for Command/timeouts:
Artifacts:
Publication status: PR #692 updated here; bead notes update follows. I will not inspect the live runner DB; monitoring uses text log/process state only until the DB is stopped/copied. |
|
SQLite PR #692 audit update for Command/timeouts:
What happened:
Stopped DB copy counts (
Artifacts:
Publication status: PR #692 updated here; bead notes update follows. Next step: isolate why the runner control DB stays locked beyond the 10s upstream busy timeout. I will not retry full |
|
Diagnostic update for the SQLite project-unit
These rows are not final SQLite pass/fail results because the suite did not complete. Current blocker is again the host-runtime wait/deadlock wedge shape, not a final SQLite failure list. |
SQLite audit update: broad fcntl diagnostic hit process-worker OOBThis was a bounded diagnostic run, not a completed SQLite pass/fail result. Command and timeout policy:
Observed blocker:
Stopped-copy counts from copied
Diagnostic failed rows before invalidation, not final SQLite failures:
Artifacts:
Scope/skips status:
Publication status: PR milestone comment published here; no new commits pushed for this diagnostic yet. Next step is to isolate whether the |
|
kd-nbh resumed the full SQLite project-unit audit from the post-OOB-fix PR head. State:
Next visible milestones:
I will preserve exact commands, timeout policy, pass/fail/omit/running/job/case counts, incomplete reasons, and artifacts. I will not claim tests pass unless a selected target actually completes." |
|
kd-nbh PR #692 progress checkpoint (2026-06-19T07:15Z): resumed from post-OOB-fix head Setup is complete enough to start the full SQLite project-unit run: repo build, zlib, SQLite CLI, Tcl, and Fresh Explain counts:
No pass/fail claim for the full suite yet. Starting the full Node execution next, then browser after Node reaches a terminal result. |
|
kd-nbh full Node run started (2026-06-19T07:17Z). Command/artifacts: I am not claiming pass/fail yet; this comment is just the visible start marker for the full Node target. |
PR #692 kd-nbh Node full run checkpoint - incompleteStatus: incomplete platform/runtime crash-wedge, not a SQLite pass/fail result and not a test-pass claim. The runner was stopped after command.log stopped changing and the Node process was sleeping at 0% CPU. Wrapper exit code after SIGTERM: 143. Worktree: Commit: Command: timeout 86400s scripts/run-sqlite-project-unit-tests.sh --host node --permutation all --jobs 32 --timeout-ms 86400000 --results-root test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-post-oobfix-20260619T071553Z --workdir /Users/brandon/src/kandelo-gascity/worktrees/kandelo/kd-nbh-pr692-full-run/test-runs/sqlite-project-unit-all/kd-nbh-all-node-jobs32-post-oobfix-20260619T071553Z/workdir --keep-workdirEnvironment highlights:
Counts from copied stopped
Total jobs: 10,353. Recorded cases before stop: 2,993. Recorded case errors before stop: 264. Remaining unrun jobs: 10,011 ready plus 32 running at termination. Omissions/skips: harness omit count is 0. Completed job output includes SQLite upstream warning Failure cause buckets from the copied DB:
Command-log crash signatures before termination:
First trap lines: Last visible runner progress lines: Artifacts:
Browser status: fresh post-OOB-fix explain completed earlier with browser total_jobs=10,366, done=0, failed=0, omitted=0, running=0, ready=10,366, cases=0, case_errors=0. Full browser execution has not started yet in this resumed session; starting it next unless this Node platform crash-wedge is declared blocking for browser. Complete failed-job catalog grouped by likely cause (output excerpts are in allocation-abort
cannot-determine-platform
database-locked
expected-got-mismatch
other-sqlite-error
readonly/permissions-mismatch
Running jobs at termination:
|
PR #692 kd-nbh browser full run startedAfter the Node run was published as incomplete at #692 (comment), I started the full browser target from the dedicated worktree. Command: Environment: PR head Artifacts: |
PR #692 kd-nbh browser active-run checkpointThe full browser target is still running, not complete and not a pass claim. Run root: Latest log-only progress sample: I am not reading live |
PR #692 kd-nbh browser 30-minute checkpointBrowser full target is still running. This is not complete and not a pass claim. Run root: Still not reading live |
2 participants
Summary
Adopts the SQLite project-unit harness work from PR #5 into Kandelo and records
both-host validation status against SQLite's official
fullpermutation.This PR adds
scripts/run-sqlite-project-unit-tests.sh, documents the harness inthe porting guide, disables accidental default browser syscall tracing for the
SQLite demo runner, improves browser artifact snapshotting, fixes browser VFS
open-unlink lifetime behavior, and stabilizes the browser threaded-sorter path
used by
sort4.test.Validation Status
Current completion target: SQLite official
fullpermutation on both Node andbrowser. The larger
allpermutation is tracked separately askad-29m.Full hard-count report:
test-runs/gastown-sqlite-epic-synthesis/final-hard-counts.md.Node full snapshot:
/bin/bash scripts/dev-shell.sh /bin/bash scripts/run-sqlite-project-unit-tests.sh --host node --permutation full --jobs 2 --timeout-ms 21600000 --results-root test-runs/gastown-sqlite-node-full-pr5test-runs/gastown-sqlite-node-full-pr5/{command.log,host-status.tsv,node/summary.txt,node/failures.tsv,node/testrunner.db,node/testrunner.log}node 143, commandexit_status=1after summary write.kad-36gfixed the Mach-O exec-resolution wedge that caused this snapshot. No later full-suite Node DB is present in the final artifact set.Browser full snapshot:
bash scripts/run-sqlite-project-unit-tests.sh --host browser --permutation full --jobs 2 --timeout-ms 21600000 --results-root test-runs/gastown-sqlite-browser-full-pr5-snapshottest-runs/gastown-sqlite-browser-full-pr5-snapshot/{run.log,host-status.tsv,combined-summary.md,browser/summary.txt,browser/failures.tsv,browser/testrunner.db,browser/testrunner.log}browser 1, page navigation/reload while Playwright was waiting inpage.evaluate().done,failed,omit,running, andready; it does not record XFAIL/XPASS/flaky fields.Focused Superseding Results
The browser full snapshot's failed/running rows were followed by focused reruns:
sysfault.testwritecrash.testkad-wtb.19: browser executable visibility/materialization after repeated crash-child iterations.writecrash.testwalfault.testkad-wtb.20: browser Tcl abort plus kernelmunmaptrap.walfault.testlike.testlike-14.2concern is timing-threshold behavior.like.testdefault caplike-14.1, notlike-14.2.savepoint6.testsort4.testkad-wtb.9.sort4.testkad-wtb.21: Node temp database open failures insort4-2.3/2.4/2.5/2.6/2.8.No SQLite test was skipped or xfailed as a substitute for runtime/platform work.
Artifacts
test-runs/gastown-sqlite-node-full-pr5/test-runs/gastown-sqlite-browser-full-pr5-snapshot/test-runs/gastown-sqlite-epic-synthesis/summary.mdtest-runs/gastown-sqlite-epic-synthesis/final-hard-counts.mdtest-runs/kad-wtb13-like-*test-runs/gastown-sqlite-browser-full-pr5-snapshot/browser-fault-recheck.mdTest Verification
Latest child branches recorded the full Kandelo gate suite before merge into
integration/kad-wtb-sqlite-testing:cargo test -p kandelo --target aarch64-apple-darwin --lib,cd host && npx vitest run,scripts/run-libc-tests.sh,scripts/run-posix-tests.sh, andscripts/dev-shell.sh bash scripts/check-abi-version.sh.