shape-doctor.py: an own-copy validator row for a root that lost scripts/repo_shape.py names the validator as its NEXT, so a reader whose Python reaches another repo_shape sees the pasted line pass; the NEXT becomes the cure, git checkout HEAD -- scripts/repo_shape.py #697

Goal: when an own-copy validator crashes because the root lost scripts/repo_shape.py, the row's NEXT puts the file back. A reader who pastes that line repairs the root instead of re-running a test that can pass on their machine.

Filed by lane openRepoShape-3 on Brett Heap's words of 2026-10-16T12:24:41Z ("give #460 to lane 3") and 13:11:50Z ("C, and no objection to #678"), option C of lane 3's question on #460's L26.

What happens (measured on main d4a7fab by lane 3's triage):

• A root that lost scripts/repo_shape.py reads DRIFTED (1 copy-missing). Its manifest row says FINDING ... crashed ... ModuleNotFoundError: No module named 'repo_shape'; scripts/repo_shape.py, which it imports from beside itself, is not there. Its NEXT is python3 <root>/scripts/validate-manifest.py --root <root>.
• Since shape-doctor.py: every validator child starts with -S, so no .pth in the user's or a virtualenv's site-packages stands in for the repo_shape.py a root lost, and the manifest row is the gate's own ModuleNotFoundError (#342) #636 the doctor runs each validator child with -S and no inherited PYTHONPATH. A reader whose own PYTHONPATH, or a .pth in whose site-packages, reaches a repo_shape runs the pasted NEXT and sees manifest ok, exit 0.
  • python3 -S stops the .pth route, but not the PYTHONPATH one.
  • Only -E -S stops both, and option B, which added them, was declined.
• The verdict line and the shape currency row still name the missing file, so nothing is hidden. But the one line a reader runs contradicts the row.

Where it is (shape-doctor.py, line numbers from d4a7fab):

• crashed() reads a validator's crash (about :1393), and validator_row builds the row from it (about :1834-1882).
• The own-copy rows that call them: manifest, pins, and any other row that runs a project's own copy of a validator.
• validator_command, and its PowerShell spelling (shape-doctor.py validator_row falls back to the standard's TEMPLATE validator when a root carries no copy of its own, and prints that path as the NEXT command without the PYTHONPATH it runs it under, so pasting the manifest, pins or family row's NEXT dies with ModuleNotFoundError: No module named 'repo_shape' and exit 1, the code a validator uses for a finding #238), print today's NEXT.
• The pattern to follow is the unloaded-manifest row's cure: git -C <root> checkout HEAD -- <relative> (shape-doctor.py: for a root whose project.yaml does not load AND whose scripts/repo_shape.py is gone, the manifest row's NEXT promises its validator "refuses it the same way until it loads", but pasted it dies on ModuleNotFoundError: No module named 'repo_shape', exit 1, so the person is sent to a command that cannot say what the row quoted #298), at about :1983-1993. It is :1991 on main c6122e8.

The change (option C):

• Where an own-copy validator row's crash is the root's lost scripts/repo_shape.py, the case the row's reason already names, its NEXT is git -C <root> checkout HEAD -- scripts/repo_shape.py.
  • Quote it as the unloaded-manifest row quotes its cure, in both dialects.
  • The WHY says to re-run the validator after it, and names that command.
• Where HEAD no longer holds scripts/repo_shape.py, there is no NEXT, and the WHY says why.
• Every other NEXT stays the project's own command. No -E or -S is added anywhere.
• The verdict and the exit code do not move.

Tests:

• A root that lost scripts/repo_shape.py:
  • each own-copy row that crashes on it gets the cure as its NEXT;
  • pasted, the cure restores the file, and the validator then passes.
• A root whose HEAD lacks the file gets no NEXT.
• A crash with another cause keeps the validator's command as its NEXT.
• The new NEXT passes the NEXT-is-a-command guards in tests/test_doctor_next_is_a_command.py, including the comment-tail guard (tests/test_doctor_next_is_a_command.py: the comment-tail guard reads each string constant alone and wants a blank after #, so a tail split across +, one led by a formatted value, or #x would reach a NEXT unseen #679).

From #460, #460 (comment) (L26, R342).

🤖 Generated with Claude Code