Repository navigation
check_fleet_pv_shapes_gate reds guard-tree when the fleet pv is older than a shape key the tree introduces (allowEmpty, car 0.70.0) #4456
Description
guard-tree on #4429 (run 36195202164, job 108270095631, intel-clean-room-4) FAILs check_fleet_pv_shapes_gate.sh with reason=pv-no-verdict. The fleet-pinned pv 0.69.1 exits 3 with error: shape release-readiness-v1.refusal uses unsupported allowEmpty.
allowEmpty (#3610) arrives in the same car as the first contract that uses it. No released pv has it: not v0.69.1, not v0.69.5-rc.1, not main. So no fleet update can clear this before 0.70.0 ships, and the guard classes pv exit 3 as a possible tree defect. This is the 0.65.2 incapable condition one release on: fleet state, which a required gate may not red on.
Fix: pv exit 3 with its parse refusal uses unsupported <key> becomes UNMEASURED reason=incapable, but only when this tree's own ontology/shapes.rs parses that key. A typo and any other exit 3 stay RED. The fix adds self-test rows and a mutant, and was checked against the real lambda pv 0.69.5-rc.1.
🤖 Generated with Claude Code
Activity
fix(guard): fleet pv shapes gate reads only NODE_KEYS/PROPERTY_KEYS a…
pvbin="$(command -v pv)") — on a fleet-clobbered runner (pv 0.65.2, no census) it reads 'no consumer' and REDs an honest anchored rise; check_apr_bin_pinned.sh is blind to the form #3679pvbin="$(command -v pv)") — on a fleet-clobbered runner (pv 0.65.2, no census) it reads 'no consumer' and REDs an honest anchored rise; check_apr_bin_pinned.sh is blind to the form #3679noahgift commentedon Oct 12, 2026
evidence for close: 989cb01 (#4655), ancestor of main 6bd89b0. Checked at main: scripts/check_fleet_pv_shapes_advisory.sh reports pv exit 3 "uses unsupported " as UNMEASURED reason=incapable only when this tree's own pv parses that key; its self-test covers the typo, other exit-3 and numeric cases, with a mutant.
noahgift commentedon Oct 12, 2026
Cop batch close (FLOW-004 drain, ≤43/d): fix verified on main — merged #4655 ; ancestor of origin/main: 989cb01 . Evidence: #4456 (comment)