Beagle SCM

CODE-022: put.js hashlet resolver — dead-ternary bug + fold to resolveHexAny

Deps: none to start. The BROADER hashlet-resolver dedup is owned by CODE-020 (resolveHashlet reconcile → shared/resolve.js); this is the put.js spin-off CODE-020 blockers explicitly told to file separately. A real bug, not just tidiness — carries its own repro. See resolve-hash.

Input

Context

• put.js:342 resolveHex is a stale copy of core/resolve.jsresolveHex.
• put.js:343 return k.getObject(hexish) ? hexish : hexish — BOTH arms return hexish, so a NON-EXISTENT full sha is accepted as valid.
• The copy drops the eachRemote prefix fallback and never reaches k.resolveHexAny, so a mid-history commit hashlet silently misses.
• Put's #sha/#hashlet target resolution rides exactly this function.

Goals

• Fix the dead ternary: an absent full sha resolves to nothing, not itself.
• Route put's hashlet→sha through k.resolveHexAny (any-object reach) or resolve-hash, retiring the local resolveHex copy.

Constraints

• Behavior-preserving EXCEPT the bug fix; repro-first (abc policy).
• be/ JS only, drive via jab; the fold joins CODE-020's resolver group.

WIP

Design decisions

• The canonical reach is k.resolveHexAny (store hashlet60 index): full sha verbatim iff present, 6..40 prefix unique-or-nothing, all objects.

TODOs

• repro: put ...#<absent-full-sha> is wrongly accepted today
• repro: put ...#<mid-history-hashlet> silently misses today
• fix put.js:343 ternary — absent object → fail/undefined, not pass-through
• fold put.js:342 resolveHex → k.resolveHexAny (drop the local copy)
• confirm banner target (put.js:336) composes via URI.make (cluster C)

Blockers and bummers

• Reconciling onto one resolver is CODE-020's repro-gated group; land the put.js bug fix independently, then let the fold ride that group.

Outcome

• Not started. Suggested commit: CODE-022: fix put hashlet dead ternary + resolveHexAny.