family.py add: after #646's move back of a member's gitlink, a .gitmodules that a git mv made name the variant is not mended, a file or link in the way has no remedy that keeps the member, and the plain mv it prints orphans a live submodule checkout #761

Goal: every command an add refusal prints for a member recorded in the wrong case can be typed as printed, in each state git can leave it in, and leaves the member registered, its .gitmodules section naming its row's path, and any live checkout openable.

Filed by lane openRepoShape-5, from #646's PR (#670, head a837a023: its checklist candidates (e), (g) and (h), "Left out", and RV670's N1 from lane 3), and from the checklist candidate of #681's PR (#747, "family.py's case-variant move remedy breaks a live submodule checkout"). The validator halves of (e) and (g) are in the sibling issue for validate-family.py; this one is scripts/family.py only.

Lands after: lane 5's #646 (PR #670, fix/646-add-member-variant-remedy), which adds the RESTORE remedy these lines are about, and #685 (family.py add, #574's spelling remedies: <new name>, a pulled Members/README, git mv --sparse), which edits the same function. Line numbers are at #670's head a837a023 where the code is new, and at main 66b230ed for MOVE and MOVE_BACK.

Lines

• (e) After git mv members/<P> Members/<P>, the printed move back leaves .gitmodules naming the variant. git mv rewrites the section's path to Members/<P>. The index-only move (scripts/family.py:3083-3111 at a837a023: git rm -r --cached --ignore-unmatch -- <variants>, then git update-index --add --cacheinfo 160000,<commit>,<row>) puts the gitlink back at members/<P> and leaves the section where it is. Measured by w646b (gitmv-head.log): git submodule update --init then fails No url found for submodule path 'members/Atlas', the gate refuses member-uninitialized, and the closing mv (below) would move the member's only checkout out of the holder. Lane 3 asked for this in RV670's N1 ("Get (e) onto Tracking: setup, scaffold, family and the shared helpers, creating projects and families #459, and name that state"). Direction: where a section of .gitmodules gives a variant's path, print git config -f .gitmodules submodule.<name>.path <row path> and git add .gitmodules beside the two lines, and commit .gitmodules in the same commit. _submodule_paths (scripts/family.py:3096 at main) already reads each section's name, path and url.
• (g) A file the index holds at or under the row's path, beside the variant's gitlink, has no remedy that keeps both (_members_spellings, scripts/family.py:2917-2983 at a837a023: such a variant is "a second gitlink" and keeps family.py: add refuses, before the member is read and on every disk, any path under another spelling of members/, in its index, in HEAD or on disk, and names a remedy per entry that never moves the members' checkouts (#362, #363, #364) #574's remedy, which takes the variant out and un-registers the member). Measured by rv670b for a file AT the row's path: typed, it leaves HEAD with 100644 blob ... members/IRRS, the gate refuses member-gitlink-absent (exit 2), and add of another member then goes through (exit 0). A file UNDER the row's path was not run. The state predates family.py: add's refusal of a member's own gitlink held only under another spelling of its row's path moves it back to that path in the index alone, so the printed remedy no longer takes the member's gitlink away from its row (#646) #670. Direction: say what the person must settle first, since neither the move back nor family.py: add refuses, before the member is read and on every disk, any path under another spelling of members/, in its index, in HEAD or on disk, and names a remedy per entry that never moves the members' checkouts (#362, #363, #364) #574's remedy keeps the member and the file: the file goes aside with a committed git mv, as _members_is_a_file already says for members itself (scripts/family.py:3162-3190 at main), and then the move back. Prove each order on a probe holder before the text says it.
• (h) A file or a link the index holds at members itself is never named. Measured by rv670b, typed line by line: with every member at a variant and members a file in the index, add prints the move back; git rm takes both variants out (exit 0), and each git update-index --add --cacheinfo then fails fatal: ... 'members/IRRS' appears as both a file and as a directory (exit 128). The next add prints the same move again, now from HEAD. A committed link left by a hand rename (members -> Members) behaves the same. family.py: add refuses, before the member is read and on every disk, any path under another spelling of members/, in its index, in HEAD or on disk, and names a remedy per entry that never moves the members' checkouts (#362, #363, #364) #574's remedy also un-registered the member there, so it is not a regression; the rule asks about entries at or under a row's path, not above it (:2917-2919). Direction: where the index holds an entry at members, name it first, with the same file-aside move as the line above, and print the move back after it; _members_is_a_file already has the wording for the mount path.
• The plain mv the refusal prints orphans a live submodule checkout. Read from source, not measured here. At a837a023 the closing text prints, for a variant directory left on a case-sensitive disk, mv -- <variant> ../<variant with / as -> (scripts/family.py:3113-3126); at main the MOVE remedy prints mv -- <variant> <new name> for "what is on disk and untracked" (:3033-3037, :3074-3075) and MOVE_BACK prints mv -- <members> members (:3076-3081). A hand-mounted checkout's .git file names its store by a relative path (gitdir: ../../.git/modules/Members/<P>), and its module config keeps core.worktree relative too. validate-family.py: a staged git rm is named again and fails retyped, the member-gitlink-absent remedy ends in a refused add, a shape-pin row with no path: reads None, and #639's mv orphans a live checkout and is stated by OS #681's N-1 measured the same mv for the validator with plain git (git submodule add ... Members/IRQS, mv -- Members/IRQS ../Members-IRQS, then git -C ../Members-IRQS status says fatal: not a git repository; re-pointing the .git file alone says fatal: cannot chdir to '../../../../Members/IRQS'). Direction: validate-family.py and validate-pins.py: a staged gitlink removal says to commit it, member-gitlink-absent puts the gitlink back by hand, the variant mv is guarded, and a pin row with no path: is shape-pin-row (#681) #747 already gave the validator the words and the guard (_wrong_case_move, templates/family-root/scripts/validate-family.py:577-639 at 031ba75e): [ <variant> -ef <row path> ] || [ -e <variant>/.git ] || mv -- <variant> ../<new name>, then where the directory holds a .git file, a live checkout, where its commits are kept and to delete it by hand once nothing only it holds is needed. Use the same text here, and decide the <new name> placeholder with family.py add: #574's spelling remedy gives a generic <new name> after git mv members Members, git rm --cached for a pulled Members/README, and a git mv --sparse that needs git 2.34 or later #685's first line. Prove each step on a probe holder.

Verify: tests/test_family_add_member_variant.py (new in #670; 37 tests that type every printed line and then the gate), tests/test_family_members_case_variants.py and tests/test_family_add_refusals.py pass on Linux. One new case per line, typed as printed to the gate and then to the add that was refused: a git mv members/IRRS Members/IRRS holder (the gate exits 0, the section says members/IRRS, git submodule update --init succeeds); a file at the row's path; a file under it; a file at members; and a live checkout at the variant whose commits survive the printed step. No macOS or Windows run (CI is Linux only, #597): the folding disk is emulated by disk_as, as #670 does.

Overlaps. Open branches that edit scripts/family.py: fix/646-add-member-variant-remedy (#670, _spellings_refusal), fix/684-family-holder-states (70153862), fix/686-bump-reachability (7567e60e), fix/688-family-init-values (aa95b6b2), fix/689-next-push-refspec (fa513848), fix/692-raw-folder-paths (#731, ab55c29f) and fix/693-local-remote-dir (96ebe450); by the file lists none of the others edits _members_spellings or _spellings_refusal, but #684 lists the case-variant git mv remedy among its lines, so read its head before starting. fix/681-validate-family-remedies (#747) and fix/648-bootstrap-and-gate-order (#682) edit the validator, not this file. scripts/family.py is not a pinned shape copy, so no upstream-changed row follows.

Parent: #459.

Milestone: v1.1.1. Each state needs a hand rename of members/, a git mv of a member's mount, or a file where a member goes: a follow-up to #646, not a regression, and it blocks nothing in v1.1.0.

🤖 Generated with Claude Code