Beagle SCM

POST-022: submodule-recursive post dumps whole sub history, per-line label

be post be://localhost/beagle (and be post ssh://…) recursing into the abc submodule prints the submodule's ENTIRE commit history — ~150 post abc/?<sha>#<msg> rows — each preceded by a bare abc label line on its own, producing pages of unreadable, wrecked output. It should emit only the commits actually being posted/pushed (the delta vs the remote), through the shared ROWS banner, with a compact per-row abc/ prefix — not a standalone abc line per entry, and not a full-history replay. Distinct from POST-021 (the BNOROOM push failure); follows POST-020/POST-019 (git-push output via ROWS / gitlink banner).

Input

Context

Two faults in the submodule-recursion output of be post <remote>.

• Repro: be post be://localhost/beagle from /home/gritzko/beagle.new (the abc

  submodule is behind its remote) → ~150 lines, alternating a bare abc label line and a post abc/?<sha>#<msg> row, back to 02 Mar bf378f3a be round-trip fixes.

• Fault A — every row is preceded by a STANDALONE abc line (the submodule label

  leaking per-row instead of one compact prefix / ROWS scope).

• Fault B — the FULL history is enumerated, not the delta: post: 0 change(s) is

  printed yet every historical commit is listed, so it replays history it is not committing. Likely the same sub history walk as POST-021.

Goals

Submodule-recursive post shows only the posted/pushed delta, compactly.

• Route sub rows through dog/ROWS with a compact abc/ prefix; no bare abc line

  per row (BE-005, like POST-018/POST-020).

• Emit only the delta being pushed, not the whole sub history.
• Repro test: a post recursing a behind submodule asserts bounded, prefixed output

  (no full-history dump, no per-line label).

Constraints

• ABC idioms; route status through ROWS; errors stay (BE-003). ssh/git uses --nosub.

WIP

Design decisions

(pending)

TODOs

• Locate the per-row abc label emitter (the submodule-recursion output wrapper;

  cf. POST-020's keeper notice routing through dog/ROWS).

• Locate why the FULL sub history is enumerated vs the delta (couples to the

  wpush_walk_commit history walk in keeper, POST-021).

• Repro-first (CLAUDE.md §17): a test asserting bounded, prefixed sub output.

Blockers and bummers

• Couples to POST-021 (same submodule commit-history walk); coordinate the fixes.

Outcome

• (open)