update-shape.py: update-accept-local-missing-name asks only the copies this run's --accept-local names, so a copy an earlier apply kept (read unchanged after) is never asked, and the next apply writes a siblings.py that cannot import from it #1082

Goal: update-shape.py apply refuses, before any write, every run that would leave a copy importing a name from a copy that does not define it, wherever the kept copy came from: this run's --accept-local, or an earlier run's.

State

• _refuse_kept_copies_missing_names (update-shape.py apply --accept-local scripts/repo_shape.py can keep an old helper that a newly copied file imports from, and make bootstrap then fails with ImportError #536, update-shape.py apply --accept-local: a kept .py copy that imports a name the run's new repo_shape.py dropped passes, and make bootstrap then dies on an ImportError #604) builds its kept set from the rows row.state in (BOTH, LOCALLY_MODIFIED) and row.path in accepted (update-shape.py:4288-4289 on main 4a6b9afd). So it asks only the copies this run's --accept-local names.
• A copy an earlier apply kept reads unchanged in the next check. The next apply never asks it, and writes a new importer beside it.
• Its docstring calls it "the general guard" (:4210).

What happens (measured)

rv1075, lane 5's independent reviewer of #1075 (#1055), measured this offline.

• Setup: a holder cut at main 6476c170, whose scripts/bootstrap.py has no untraced, with its own edit to scripts/bootstrap.py committed.
• First apply: update-shape.py apply --at 00f6d7f7 --accept-local scripts/bootstrap.py exits 0. That target's siblings.py imports no untraced, so nothing is refused.
• Second apply: update-shape.py apply --at c3cf5154 (the family root's siblings.py: with a token in use, the clone and the fetch that carry it run in bootstrap.py's untraced() environment, so make siblings and family.py siblings write the token to no trace the caller asks git for (#1055) #1075's head), with no --accept-local.
  • check reads unchanged scripts/bootstrap.py and upstream-changed scripts/siblings.py.
  • The apply exits 0 and prints copied scripts/siblings.py. The validators pass, and it commits.
  • Then scripts/siblings.py --help fails with ImportError: cannot import name 'untraced' from 'bootstrap'. So would make siblings, park and resume.
• The one-step run is refused, as update-shape.py apply --accept-local scripts/repo_shape.py can keep an old helper that a newly copied file imports from, and make bootstrap then fails with ImportError #536 says. apply --at c3cf5154 --accept-local scripts/bootstrap.py exits 2 with scripts/bootstrap.py (kept) lacks untraced, which scripts/siblings.py imports from it. The holder's files, HEAD and status are unchanged.
• On main already (by the same reading, not measured): the family root's bootstrap.py and siblings.py: under GNU make 4.4's named-pipe jobserver, a member's make that says it is 4.4 or later is handed the holder's MAKEFLAGS as it is, so it runs its jobs together again; every other make still gets no jobserver and no -j (#987) #1023's member_make_environment has the same gap. A holder that kept, by an earlier apply, a bootstrap.py from before the family root's bootstrap.py and siblings.py: under GNU make 4.4's named-pipe jobserver, a member's make that says it is 4.4 or later is handed the holder's MAKEFLAGS as it is, so it runs its jobs together again; every other make still gets no jobserver and no -j (#987) #1023 is written main's siblings.py, which imports that name.

Direction

• Ask every copy this run writes against the copy it imports from, as the run leaves that copy, not only against the copies this run's --accept-local names. The copy it imports from may be:
  • written by this run;
  • kept by this run's --accept-local;
  • left on disk as an earlier run kept it.
• Keep the refusal's code, update-accept-local-missing-name.
• Where the kept copy reads unchanged, the line could say an earlier run kept it. That is the writer's call.

Verify

• The two applies above: the second is refused before any write, naming scripts/bootstrap.py and untraced. The holder's files, HEAD and status are unchanged.
• The one-step refusal is unchanged.
• The same for member_make_environment, with a holder cut before the family root's bootstrap.py and siblings.py: under GNU make 4.4's named-pipe jobserver, a member's make that says it is 4.4 or later is handed the holder's MAKEFLAGS as it is, so it runs its jobs together again; every other make still gets no jobserver and no -j (#987) #1023.

Found by lane 5's independent review of #1075 (#1055), rv1075 (#1075 issuecomment-6302090874).

Milestone: v1.2.0 (placement is lane 3's call).

Lane: openreposhape-5 (openRepoShape-5)