Beagle SCM

STATUS-012: divergence row says miss for a commit already PATCHed into the wt

Incident 2026-07-17, work/PUT-010: after jab patch ?e3a937bd absorbed the behind-commit (wtlog patch row live, wt full of pat/mrg rows), status still lists it as miss ?e3a937bd#PATCH-013: fix timestamping. miss reads as "you don't have this work" — wrong once the wt carries the absorbed bytes pending post. The divergence walk (status.js computeDivergence) should consult the wtlog's in-scope patch rows (PATCH ancestor-skip: parent ∪ foster reachability) and mark an absorbed commit distinctly — e.g. pat ?<sha>#<subject> — or drop it from miss.

Input

Context

• The behind-set is computed against the track tip; absorbed-but-unposted commits are indistinguishable from truly missing ones in the current row verb.
• The absorbed sha sits in the wtlog patch row URI (DIS-030 shapes); POST will consume it — status already reads those rows elsewhere (patchStamps).

Goals

Behind-commits absorbed by a pending patch render with a distinct verb — ptch (gritzko ruling 2026-07-17, PATCH §Summary 1.1: absorbed miss commits "become ptch"; COMMIT statuses are 4-char, FILE rows stay 3-char) — leaving miss for genuinely un-absorbed work; count summary follows.

• mrg and pat file rows carry a put button in the pager, like the adv sub rows do (BE-049) — stage the absorbed file straight off the status view (gritzko, 2026-07-17).

Constraints

• Display/classification change only — no wtlog/store writes, byte-parity elsewhere.
• Ancestor-skip semantics are the truth: a commit reachable via parent ∪ foster is not miss either (verify both cases).

WIP

Design decisions

• Absorbed = behind ∩ ancestors(patchTheirs) (dag.ancestors per in-scope theirs sha) — covers ?br!'s whole replayed stack, and a NEXT/NAMED row marks exactly its one commit (its ancestors are cur-reachable, so never behind).
• Row keeps its place in the behind list; only the verb flips to ptch (4-char COMMIT status per the ruling; the pat/mrg FILE buckets stay 3-char). Summary note follows the verbs: (behind B, ptch P, ahead A), each segment omitted at 0 — byte-parity when no patch row is in scope.
• The put button rides the existing BE-041 table: ACT_PUT += {pat, mrg} in status.js; no pager change, no classify flag. The ptch ?<sha> COMMIT row carries no button (buttons live in the file-bucket loop only).
• Repro shape: the wt-operand track (get //srcwt, DIS-062). A ?branch/trunk track can't repro — patchscope.resolveOurs prefers resolveRef(cur.branch), so the advanced ref becomes ours and the patch no-ops (see bummers).

TODOs

• Repro: patch a behind-commit, assert the divergence row verb (test/status/patched, RED pre-fix).
• Fix computeDivergence to consult in-scope patch rows.
• Pin: a POSTed absorption (theirs = merge parent) is not listed at all (patched leg 3).
• pat/mrg file rows carry the put button like BE-049 adv rows (test/status/patchbtn, RED pre-fix; headless-Pager click leg: mutate + in-place refresh, no push).
• Ruling 2026-07-17 (PATCH §Summary 1.1): divergence verb is ptch, not pat — verb, summary segment and test/status/patched assertions renamed; file buckets untouched.

Blockers and bummers

• patchscope.resolveOurs resolves ours ref-FIRST (resolveRef(cur.branch) over cur.sha), so on a shared-store wt whose track ref advanced, jab patch ?<behind-sha> silently absorbs NOTHING (ours == theirs, exit 0, no row). Not this ticket's display fix — possible own ticket.
• JS post records theirs as a real merge PARENT (no foster header yet), so the parent ∪ foster pin here exercises the parent leg only.

Outcome