Beagle SCM

POST-024: a tree-URI post target is mis-slotted as a path scope — silent no-op + misleading POSTNONE

Input

Context

From work/ABC-008/abc, jab post file:../../../libdog/abc/ answered POSTNONE: no changes since — the wrong diagnosis (the wt HAD a fresh commit; the TARGET was the problem).

• file:<rel> carries a scheme but NO authority, so post's slot parse (verbs/post/post.js parseSlots, DIS-054 Host>Query>Path) reads its PATH as a commit path SCOPE, not a target.
• The ../… scope escapes the wt, silently matches nothing, and the empty-commit gate (post.js ~647) throws POSTNONE: no changes since base — user-hostile misdirection.
• Wire push (GIT-013) serves ssh:/http(s): Host slots only; a local file:/be:// tree target has NO leg at all.

Goals

• a scheme'd arg post cannot serve refuses LOUDLY with its own code (BE-003 spirit: one uniform line naming the unsupported target), never silently re-slots as a path scope.
• a ..-climbing path-scope arg throws NAVESCAPE, never a silent empty scope.
• (design, cross-link PATCH-011) the wt→orig merge flow goes through patch-with-fetch, NOT post: post to a local tree target stays refused.

Constraints

• URI classification via uri._parse / the URI class only (work bans regex/manual URI parsing); minimal diff in verbs/post/post.js.
• repro test first; existing tests untouched.

WIP

Design decisions

• POSTSCHEME gate fires only on scheme'd args with TARGET shape (authority/query/path-slot path) — scheme-lexing message words (T:v2, POST-024: msg) stay prose, pinned by a control case.
• be://host refuses BEFORE the GIT-013 Host-slot branch; narrow's value stays byte-identical (resolveInTree validates, never rewrites).

TODOs

• repro test test/post/target-refuse (red-first on the exact ticket bug; POSTSCHEME for file:<rel>/be://…/file:/abs, NAVESCAPE for the climb, store tip intact, colon-message control)
• slot parse: scheme'd non-wire arg → POSTSCHEME; climbing scope → NAVESCAPE (verbs/post/post.js parseSlots)

Blockers and bummers

• change sits UNCOMMITTED in work/POST-024, awaiting review/merge; suite there 171/172 (the 1 red = pre-existing be-js-get-oddname, JS-096 area), orchestrator re-ran the new test green.

Outcome