Beagle SCM

JS-085 HIGH: views/log/log.js::resolveHashlet re-impls store.resolveHexAny via test-only k._packs()

be log:#<hashlet> resolves a short hex to a full COMMIT sha. The log view's resolveHashlet (log.js:104) bypasses the public reader API: it reaches into the test-only k._packs(), re-mmaps every pack (already cached by the store), and re-runs a full O(all-objects) decompress-and- prefix-match scan that near-verbatim duplicates store.resolveHexAny + indexPackByWalk. commit.js already does the same job in ~6 lines. Method: Issues.

Input

Context

• log.js:109 packs = k._packs() — _packs is store.js:551

  marked "expose for tests / verification", NOT a reader verb.

• log.js:111-130 re-mmaps each pack (git.pack.mmap), walks

  every offset, inflates each object (pk.resolve), reframes the sha, prefix-matches, and re-detects ambiguity.

• This is the exact work store.js::resolveHexAny (:302) +

  indexPackByWalk (:95) already do over the cached wh128 lane.

• commit.js:101-107 does the canonical thing: k.resolveHexAny

  then keep type==="commit"||"tag"; ~6 lines, no pack re-mmap.

Goals

• Replace resolveHashlet's body with the commit.js form:

  k.resolveHexAny(hexish) then k.getObject(sha).type filter.

• be log:#<commit-hashlet> for a mid-history (non-tip) commit

  still resolves; an ambiguous prefix still resolves to NOTHING.

• Remove _packs from the production reader surface (or fence

  it test-only) so views cannot re-enter the pack walk.

Constraints

• Byte-equivalent log output: the existing log parity tests

  (test/js/log/*) must stay green — same resolved sha, same ambiguity → empty behaviour as KEEPResolveHex.

• Use only the reader's public verbs (resolveHexAny, getObject,

  resolve.resolveHex tip fast-path); no git.pack in log.js.

WIP

Design decisions

• Mirror commit.js:101-107 exactly so the two hashlet resolvers

  converge on the ONE store.resolveHexAny path (see JS-090).

• log keeps the tip fast-path resolve.resolveHex(k,hexish)

  first only if it changes no bytes; else fold into resolveHexAny.

TODOs

• Repro-first: a test/js/log case log:#<mid-history-hashlet>

  + an ambiguous-prefix case asserting empty, against current bytes.

• Rewrite resolveHashlet over k.resolveHexAny + getObject type.
• Drop _packs from reader (store.js:555) or fence test-only.

Blockers and bummers

• Pairs with JS-090 (three overlapping short-hex resolvers).

• 2026-07-07 audit: STILL OPEN (line refs shifted) — views/log/log.js:121 resolveHashlet

  still calls k._packs() (:126); shared/store.js:557 still exports _packs; the commit.js canonical form (k.resolveHexAny+getObject) is not adopted in log.js.

Outcome

• (pending)