Twinned from https://github.com/cfsh/ninefold/issues/850 by tools/beads/import_github.py's reconcile pass.
Suggestions from the ninefold-reviewer review of #847 (the post-merge UI gate goes off by default). Neither belongs in that PR — one is a different tool, the other is a design-doc amendment that design_docs/README.md requires be its own PR.
1. tools/office/spawn_dev.py still requires Godot to call a box merge-capable.
Box.can_merge is dotnet and godot, on a premise this PR falsifies:
> merge_pr.py runs BOTH ./presubmit.py --all and probe_gate.py. — spawn_dev.py:518
After this change an ordinary merge runs neither probe_gate.py nor anything else needing an engine — ALL_GATING in presubmit.py is dotnet/python/shellcheck. So a box with the SDK and no Godot now gets the kickoff text at spawn_dev.py:589:
> This box has dotnet but no Godot: ./presubmit.py --all runs {lanes} here, but probe_gate.py cannot, so you cannot merge.
That constraint no longer exists. Either relax the predicate (with the gate_note prose and test_spawn_dev.py:1034), or keep it deliberately and restate the reason as "you could not sweep what you merged" rather than "the merge runs probe_gate.py". It errs conservative, so nothing is broken today — but the stated reason is now false, which is how the next reader gets it wrong.
2. design 005 S12 says the full post-merge sweep is permanent; nothing in the doc records that it is currently off.
> S12 — Keep the full post-merge sweep permanently, the only thing that catches #387-class coupling. (design_docs/005-probe-selection-and-build-config.md:134)
The PR body handles this honestly: it names S12, calls the change a temporary deviation rather than a supersession, and says re-enable once S4/S4b land. But that record lives only in a PR body and a code comment, and design_docs/README.md is explicit about why that is not enough:
> A correction living only on the implementation issue leaves the doc quietly misleading anyone who reads it the way this file tells them to.
Issue #686 (priority:next) implements 005 from the doc as a self-contained artifact. Owed: an Amendments entry on 005 recording the deviation and the condition for reverting it — one PR of its own, separate from the code that motivated it, per that README, and it waits for @aedanpope like every design_docs/** change. A comment on #686 is the cheap half and worth doing either way, so whoever restores selection knows to flip RUN_PROBE_GATE back to a default.
None.
No comments.