From the round-2 review of #265 (follows #268).
Where: scripts/fleet/check_cards.py main() closing line ("AUDIT: fix site_audit.json."), and _fleet/README.md's remedy list.
AUDIT fires for a missing or non-integer figure_at_pin and for a bad pin time. Both fields are derived by build_site_audit.py. A maintainer following the remedy types "figure_at_pin": <card> by hand. The provenance test only requires it to equal the card, so it passes, and WRONG then compares the card with itself: the failure #231 and #268 exist to prevent. The same output says "never edit it by hand" about that field. The README's WRONG bullet also still opens by calling figure_at_pin mistyped.
Fix: AUDIT's remedy should say to re-derive the audit with build_site_audit.py against a snapshot at its pins. Reword the WRONG bullet's opening.
From the round-2 review of #265 (follows #268).
Where:
scripts/fleet/check_cards.pymain()closing line ("AUDIT: fix site_audit.json."), and_fleet/README.md's remedy list.AUDIT fires for a missing or non-integer
figure_at_pinand for a bad pin time. Both fields are derived bybuild_site_audit.py. A maintainer following the remedy types"figure_at_pin": <card>by hand. The provenance test only requires it to equal the card, so it passes, and WRONG then compares the card with itself: the failure #231 and #268 exist to prevent. The same output says "never edit it by hand" about that field. The README's WRONG bullet also still opens by callingfigure_at_pinmistyped.Fix: AUDIT's remedy should say to re-derive the audit with
build_site_audit.pyagainst a snapshot at its pins. Reword the WRONG bullet's opening.