JAB-005: split js/ (jabc) out of beagle into its own project on libabc + libdog

The beagle tree splits into subprojects; js/ — jabc, the JavaScriptCore runtime building the jab binary — becomes a separate project consuming libabc and libdog (DOG-007) as submodules. JAB-001 already divorced jab from be (standalone runner, upward be/-scan), so the runtime has no build-time ties to the beagle dogs: jab links ${JSC_LIB} dog abc-core ${sodium} and nothing else. What remains is repo mechanics plus untangling the test references that reach outside js/. Blocked on DOG-007. Method: Issues.

Input

Context

js/CMakeLists.txt compiles ../abc/{POL,NET,TCP,UDP}.c straight into jab (they are stripped from abc-core), so jabc needs the abc SOURCES — the libabc submodule provides them; only the ../abc spelling changes to the in-repo path.
js/CMakeLists.txt reaches outside the dir twice: ${CMAKE_SOURCE_DIR}/test/js/get/zip.js (beagle top-level test/js/) and ${CMAKE_SOURCE_DIR}/../be/test/{ulog,store}.js (the sibling beagle-ext checkout) — the latter already breaks configure when be/ is absent, filed as JS-095; JS-061 is the umbrella for sorting JS tests to their owning repo.
beagle's root CMake gates the whole dir behind option(WITH_JS) + add_subdirectory(js); no other beagle module links against anything in js/.
be/beagle-ext executes on the jab binary; after the split, beagle-side e2e runs need a jab built from the new repo (PATH or a pinned submodule — see the T2 ruling).

Goals

A standalone jabc repo: js/ content at the root, abc + dog submodule pins, project() + JSC discovery; builds jab and runs the full JABC* ctest suite green with no sibling checkouts.
beagle drops WITH_JS + add_subdirectory(js); the full beagle ctest stays green, with jab-dependent tests moved out or pointed at an external jab.
JS-095 fixed by construction: any reference to a beagle-ext checkout is existence-guarded in BOTH add_test and set_tests_properties.

Constraints

Blocked on DOG-007: libdog must be consumable as a submodule first.
RULING needed (T2): does beagle keep jabc as a third submodule to pin a known-good jab for its e2e tests, or fully detach and take jab from PATH / -DJAB=?
The GitHub repo is created by the USER (T3); repo name suggestion jabc (binary stays jab) — confirm before creating.
abc/dog pins in jabc must track beagle's pins while be/beagle-ext straddles both trees.
Own worktree per Issues; no edits to the beagle main tree before the cut-over is approved.

WIP

Design decisions

History PRESERVED, not fresh-imported: git subtree split --prefix=js off the exported beagle main (e5071f32) → /home/gritzko/src/jabc, 96 commits, split head 14b6d109. Repo name jabc confirmed by the user 2026-07-06.
T2 resolved AS SUBMODULE: the beagle cut-over (46f87bcd) mounts jabc at js/ (WITH_JS + add_subdirectory(js) unchanged), pinning a known-good jab for beagle's e2e tests.
T2 REVERSED post-cutover to FULL DETACH: 91fd5f14 drops the js (and abc) gitlinks — beagle keeps only dog/ (libdog); no WITH_JS, no add_subdirectory(js); jab comes from -DJAB_BIN=/$JAB_BIN/PATH (be/CMakeLists.txt).
Repo name ended up jab, NOT jabc: github.com/gritzko/jab, history reforked (dd11f07 "fork off the parent repo" adds abc+dog submodules; b58be37 adds GitHub CI + folds abc INTO libdog, ABC_DIR=dog/abc).

TODOs

T1+T4b standalone: root project(jabc C CXX) dual-mode block (ABC_DIR for the stripped abc/{POL,NET,TCP,UDP}.c, guarded add_subdirectory(abc)/(dog), existence-gated JABCzip); abc server-sync 83d20aa + dog 0ff3eb7 pins; jabc main 3857f64; ctest 94/95 (RONtest only; _Alignof/_Thread_local → C23 keywords fixed in abc for the C++ TUs).
T2 ruling: submodule (see Design decisions).
T3 USER: created github.com/gritzko/jabc (2026-07-06); local repo /home/gritzko/src/jabc, origin wired.
T4a content+history: subtree split pushed to github.com/gritzko/jabc main = 14b6d109 (96 commits).
T5 beagle cut-over 46f87bcd: js/ is the jabc submodule; WITH_JS + tests unchanged; beagle ctest 386/389 (all 3 reds pre-existing/known).
T6 be/beagle-ext: published as github.com/gritzko/jab main b58be37 (fork dd11f07 + CI, pushed); be resolves jab via -DJAB_BIN=/$JAB_BIN/PATH (be/CMakeLists.txt); the abc pin is moot — abc rides inside libdog. (3857f64 superseded by the refork.)

Blockers and bummers

DOG-007 must land first. CLEARED: libdog is beagle's dog/ submodule (46f87bcd, 91fd5f14, b0fb0ce4).

Outcome

LANDED. beagle: 46f87bcd (dog/+js/ become submodules) → a2cbbe00 (repin subs) → 91fd5f14 (js/abc gitlinks dropped, full detach, libdog-only) → b0fb0ce4 (bump libdog); abc c89bb1af (C23 keywords for the C++ TUs).
jab repo: github.com/gritzko/jab main b58be37 (subtree-split history through 14b6d10, then dd11f07 fork + b58be37 GitHub CI); only submodule dog (libdog), abc sources via ABC_DIR=dog/abc.
Cancelled/superseded: the T2 submodule mount (beagle fully detached instead, jab from PATH/-DJAB_BIN=); the jabc repo name (it is jab); pre-refork heads 3857f64/14b6d109 as published mains.
Audit 2026-07-07: beagle HEAD b0fb0ce4 has no js/ tree entry, .gitmodules lists only dog; jab main tracks origin/main with no unpushed commits; console et al. verified working via ~/bin/jab.
Filed 2026-07-06, the beagle-split planning session.