Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion artifacts/findings.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -241,7 +241,7 @@ artifacts:
title: rivet 0.16.0 canonical status enum is breaking - no migration, drops 'accepted', misfits ai-found-defect
status: open
description: >
Adopting rivet 0.16.0 flips the jess store from PASS (4 warnings) on 0.15.0 to FAIL (21 errors) with no artifact changes - all 21 are the new canonical status enum (PR #419 / #352): status values accepted (x12), open (x7), resolved (x2) are no longer allowed; the enum is [draft, proposed, approved, implemented, verified, released, deprecated, rejected]. Three friction points: (1) breaking with no migration path - a previously-valid store needs 21 manual edits; (2) 'accepted' was the documented terminal lifecycle state (draft->...->verified->accepted in rivet's own release/agent guidance) and is now gone; (3) one global document-lifecycle enum doesn't fit ai-found-defect, whose natural vocabulary is open/triaged/resolved/wontfix. Suggested: ship a status remap (rivet migrate / validate --fix) + documented mapping, reconcile 'accepted', and allow per-schema status enums. jess stays on rivet 0.15.0 for now to keep the store valid and defect semantics correct. Filed https://github.com/pulseengine/rivet/issues/522 - supervising. UPDATE 2026-06-10 (iter#4): maintainer triaged and confirmed all three points as independently actionable; at their request jess added an Acceptance-Criteria block to the issue for slice 2 (re-add 'accepted' to the canonical enum - clears 12 of 21 errors, one-line schema change), with slices 1 (rivet migrate / validate --fix) and 3 (per-schema defect status enum) scoped out for separate PRs. jess stays on 0.15.0 until at least slice 2 ships. UPDATE 2026-06-10 (iter#6): slice-2 draft PR rivet#525 verified PRE-MERGE against the jess store - built pr525 head 117e9c7 and ran validate: accepted errors 13 -> 0, remaining 13 are exactly the out-of-scope slice-3 vocabulary (open x7, resolved x6 on ai-found-defect). Posted the verification table on rivet#522 (issuecomment-4673352099); PR is parked as draft only on the pulseengine.eu blog 503 process gate. jess upgrades off 0.15.0 once slice 2 ships in a release. UPDATE 2026-06-14 (iter#19): rivet#525 MERGED to main and rivet#522 CLOSED - but the merge (13:13) postdates the latest release v0.16.1 (08:20), so 'accepted' is fixed on main but NOT YET in a release. jess holds at 0.15.0 and adopts the first rivet release that carries #525; will re-run validate on that release to confirm accepted 13->0 before upgrading. UPDATE 2026-06-20 (iter#41): rivet v0.17.0 RELEASED carrying #525 (slice 2). Re-verified against the jess store on the v0.17.0 darwin release: accepted errors 0 (slice 2 confirmed shipped), remaining = EXACTLY 27 status-enum errors, ALL on ai-found-defect: 14 open + 13 resolved (slice 3, the per-schema defect status enum, NOT yet shipped); no non-status errors. Bump gated solely on slice 3. Rather than migrate 52 defect artifacts to an ill-fitting canonical mapping, jess filed the focused slice-3 ask https://github.com/pulseengine/rivet/issues/550 (per-schema status enum for ai-found-defect, or a documented defect mapping + rivet migrate). jess stays on 0.15.0 and bumps to v0.17.0 the moment defect statuses validate; offered pre-merge branch testing as for #525.
Adopting rivet 0.16.0 flips the jess store from PASS (4 warnings) on 0.15.0 to FAIL (21 errors) with no artifact changes - all 21 are the new canonical status enum (PR #419 / #352): status values accepted (x12), open (x7), resolved (x2) are no longer allowed; the enum is [draft, proposed, approved, implemented, verified, released, deprecated, rejected]. Three friction points: (1) breaking with no migration path - a previously-valid store needs 21 manual edits; (2) 'accepted' was the documented terminal lifecycle state (draft->...->verified->accepted in rivet's own release/agent guidance) and is now gone; (3) one global document-lifecycle enum doesn't fit ai-found-defect, whose natural vocabulary is open/triaged/resolved/wontfix. Suggested: ship a status remap (rivet migrate / validate --fix) + documented mapping, reconcile 'accepted', and allow per-schema status enums. jess stays on rivet 0.15.0 for now to keep the store valid and defect semantics correct. Filed https://github.com/pulseengine/rivet/issues/522 - supervising. UPDATE 2026-06-10 (iter#4): maintainer triaged and confirmed all three points as independently actionable; at their request jess added an Acceptance-Criteria block to the issue for slice 2 (re-add 'accepted' to the canonical enum - clears 12 of 21 errors, one-line schema change), with slices 1 (rivet migrate / validate --fix) and 3 (per-schema defect status enum) scoped out for separate PRs. jess stays on 0.15.0 until at least slice 2 ships. UPDATE 2026-06-10 (iter#6): slice-2 draft PR rivet#525 verified PRE-MERGE against the jess store - built pr525 head 117e9c7 and ran validate: accepted errors 13 -> 0, remaining 13 are exactly the out-of-scope slice-3 vocabulary (open x7, resolved x6 on ai-found-defect). Posted the verification table on rivet#522 (issuecomment-4673352099); PR is parked as draft only on the pulseengine.eu blog 503 process gate. jess upgrades off 0.15.0 once slice 2 ships in a release. UPDATE 2026-06-14 (iter#19): rivet#525 MERGED to main and rivet#522 CLOSED - but the merge (13:13) postdates the latest release v0.16.1 (08:20), so 'accepted' is fixed on main but NOT YET in a release. jess holds at 0.15.0 and adopts the first rivet release that carries #525; will re-run validate on that release to confirm accepted 13->0 before upgrading. UPDATE 2026-06-20 (iter#41): rivet v0.17.0 RELEASED carrying #525 (slice 2). Re-verified against the jess store on the v0.17.0 darwin release: accepted errors 0 (slice 2 confirmed shipped), remaining = EXACTLY 27 status-enum errors, ALL on ai-found-defect: 14 open + 13 resolved (slice 3, the per-schema defect status enum, NOT yet shipped); no non-status errors. Bump gated solely on slice 3. Rather than migrate 52 defect artifacts to an ill-fitting canonical mapping, jess filed the focused slice-3 ask https://github.com/pulseengine/rivet/issues/550 (per-schema status enum for ai-found-defect, or a documented defect mapping + rivet migrate). jess stays on 0.15.0 and bumps to v0.17.0 the moment defect statuses validate; offered pre-merge branch testing as for #525. UPDATE 2026-06-23 (cron loop): rivet#550 is now CLOSED (2026-06-23T08:21) - the slice-3 per-schema ai-found-defect status enum is resolved on rivet main. BUT the latest rivet RELEASE is still v0.17.0 (2026-06-19), which PREDATES the close, so no release carries the slice-3 fix yet; jess CI stays pinned RIVET_VERSION v0.15.0 and validate PASSES there (confirmed this loop with a freshly-bootstrapped rivet 0.15.0 darwin binary). CAVEAT discovered: a local rivet that had DRIFTED to 0.17.0 (build d648ce44, main 2026-06-23) still REJECTS open/resolved on ai-found-defect (28 errors, allowed = the generic [draft..accepted] enum) - i.e. even the same-day main build does not yet expose [open/triaged/resolved] for the defect schema, so the #550 resolution likely needs a schema opt-in OR my local build predates the merge. NET: bump off 0.15.0 still blocked until (a) a rivet RELEASE past the #550 fix ships AND (b) jess re-verifies open/resolved validate on it (offer pre-release verification, as for #525). OPERATIONAL: validate jess with the CI-PINNED rivet 0.15.0, NOT a drifted local 0.17.0 (the latter gives false FAILs on the pre-existing defect statuses).
tags: [release-watch, rivet, schema, breaking]
fields:
detected-by: jess loop iter#3 (2026-06-10) - rivet 0.16.0 validate on the jess store
Expand Down Expand Up @@ -796,6 +796,13 @@ artifacts:
the RC-input decode series (SBUS v1.88 + CRSF v1.89, both -> bounded RcInput).
Per-piece GREEN on synth v0.12.0, authoritative skips 3/18 all #369 unchanged,
bulk-mem clean - no new unlowerable op. Evidence: results/jess-build-falcon-v1.89.0.xml.
UPDATE 2026-06-23 (cron, falcon-v1.92.0 = DroneCAN foundation, DC-P01, relay #226):
v1.90-v1.92 series; v1.92 = "DroneCAN foundation - spar model + CanBus seam +
verified transfer layer" (the relay-side start of the CAN path, REQ-PIX-006). Tested
on synth v0.12.0: per-piece GREEN, authoritative skips 3/18 all #369 unchanged,
bulk-mem clean - the CanBus seam + transfer layer add no new unlowerable op. Also
meld v0.34.0 (resource_graph #295 / LS-A-17 hardening) - jess forward-chain fuse
stayed green. Evidence: results/jess-build-falcon-v1.92.0.xml.
tags: [release-watch, synth, fpu, miscompile, correctness, on-target, blocker]
fields:
detected-by: jess REQ-PIX-001 value-level oracle - silent miscompile resolved in v0.11.46 (GI-FPU-001, verified loud-skip exit 1); OPEN remainder is GI-FPU-002 hard-float
Expand Down
13 changes: 9 additions & 4 deletions artifacts/phase2-pixhawk.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -139,10 +139,15 @@ artifacts:
type: requirement
title: DroneCAN/CAN-FD peripheral + smart-actuator path on the 6X-RT CAN buses
status: draft
fields:
category: functional
priority: could
release: v0.8.0
description: >
jess binds the CAN path to the RT1176 silicon (the 3x CAN-FD on the 6X-RT,
2 broken out). RELAY-SIDE FOUNDATION LANDED 2026-06-23: relay falcon-v1.92.0
("DroneCAN foundation - spar model + CanBus seam + verified transfer layer",
DC-P01 / relay #226) - the portable CanBus seam + verified transfer layer that
jess will bind, the CAN analogue of the relay-hal transport matrix (DD-018).
Per-piece GREEN on synth v0.12.0 (AFD-024). jess's binding (CAN controller +
transceiver over the RT1176 FlexCAN, via the embedded-hal-async-style seam) is
the on-silicon counterpart, sequenced after the core HAL binding (REQ-PIX-012).
links:
- type: traces-to
target: FEATURE-002
Expand Down
12 changes: 12 additions & 0 deletions results/jess-build-falcon-v1.92.0.xml
Original file line number Diff line number Diff line change
@@ -0,0 +1,12 @@
<?xml version="1.0" encoding="UTF-8"?>
<testsuites>
<testsuite name="jess-build" tests="7" failures="0" skipped="0" errors="0" time="0">
<testcase name="sha256-verify" classname="jess-build" time="0"/>
<testcase name="sil-stabilization" classname="jess-build" time="0"/>
<testcase name="sil-position-hold" classname="jess-build" time="0"/>
<testcase name="meld-fuse" classname="jess-build" time="0"/>
<testcase name="loom-optimize" classname="jess-build" time="0"/>
<testcase name="synth-compile" classname="jess-build" time="0"/>
<testcase name="kiln-xruntime" classname="jess-build" time="0"/>
</testsuite>
</testsuites>
Loading