Beagle SCM

JS-095: js/CMakeLists.txt configure FAILS when the be/ submodule is absent (EXISTS-guarded test referenced unconditionally in set_tests_properties)

A be get be://... clone that lacks the be/ submodule (no be/test/ulog.js, be/test/store.js) cannot even CONFIGURE the default WITH_JS=ON build: cmake aborts with set_tests_properties Can not find test to add properties to: JABCulogLib. The add_test(JABCulogLib/JABCstore) are EXISTS-guarded (only created when be/ is present) but set_tests_properties(... JABCulogLib JABCstore ...) at js/CMakeLists.txt:117 lists them UNCONDITIONALLY. Hit while building a clone for LOG-003. Method Issues.

Input

Context

• Repro (live 2026-06-26): be get be://localhost?/beagle (clone at ?#213c9003, lacks the be/ submodule), then CC=clang CXX=clang++ cmake -DCMAKE_BUILD_TYPE=Debug -DWITH_ASAN=ON -DWITH_INET=OFF -GNinja -B. -S.. → CMake Error at js/CMakeLists.txt:117 (set_tests_properties): set_tests_properties Can not find test to add properties to: JABCulogLib → "Configuring incomplete".
• Cause: js/CMakeLists.txt:84-86 wraps add_test(NAME JABCulogLib ...) in if(EXISTS ${CMAKE_SOURCE_DIR}/../be/test/ulog.js); :87-89 wraps JABCstore in if(EXISTS .../be/test/store.js). But :117 set_tests_properties(... JABCulogLib ... JABCstore ... PROPERTIES ENVIRONMENT LSAN_OPTIONS=...) names BOTH unconditionally. Absent be/ → tests not created → set_tests_properties hard-errors.
• be/ is the JS trunk submodule; a be:// C-core clone need not carry it (the same clone also lacks be/views/log/log.js, flagged by LOG-003).
• Workaround used for LOG-003: configure with -DWITH_JS=OFF (skips the whole js/ subdir). Not a fix — it disables all JS tests.

Goals

• cmake ... -DWITH_JS=ON configures cleanly whether or not the be/ submodule is checked out; present-be/ builds still attach the LSAN suppression env to JABCulogLib/JABCstore.

Constraints

• Keep the EXISTS guards; make the set_tests_properties apply per-test only when that test exists — e.g. move the ENVIRONMENT LSAN_OPTIONS property INTO each if(EXISTS ...) block, or split the one bulk set_tests_properties so the two be/-gated names are set inside their guards. Native (full-submodule) build behavior unchanged.

WIP

Design decisions

• (open) Two clean options: (a) split line 117 — drop JABCulogLib/JABCstore from the bulk call and set their ENVIRONMENT inside the existing if(EXISTS ...) blocks at :84-89; or (b) guard the bulk call's name list with a CMake list built up only from created tests. (a) is the smaller diff.

TODOs

• Repro-first: a configure of WITH_JS=ON against a tree WITHOUT be/ (rename/hide be/ or use a be:// clone) FAILS today, PASSES after. (Fix landed inside a comment-sweep commit, no dedicated repro.)
• Drop JABCulogLib + JABCstore from the unconditional bulk set_tests_properties list (fc50e82, js/ lineage — the "# JS-095:" comment marks it; today CMakeLists.txt:143-145 of the jab runtime repo).
• Sanity: full-submodule build still sets the property — SUPERSEDED: the two tests get NO per-test LSAN env anymore; JS-064 embedded the JSC suppression in the jab binary itself, making the external env redundant.

Blockers and bummers

• Blocks any C-core-only be:// clone from building the default WITH_JS=ON config; forces -DWITH_JS=OFF and loses all JS test coverage in such trees.

Outcome

• LANDED fc50e82 (js/ lineage, 2026-07-03; folded into "DOG-006: all comments support inline StrictMark"): the bulk set_tests_properties no longer names the be/-gated JABCulogLib/JABCstore, so configure succeeds without be/. Verified 2026-07-07 in the jab runtime repo CMakeLists.txt (:110-114 EXISTS-guarded add_test; :143-145 bulk list excludes them, "# JS-095:" comment).
• OVERTAKEN by the restructure too: js/ left the beagle build entirely (JAB-005 46f87bcd, dog/+js/ became submodules; the runtime forked to its own repo dd11f07), so beagle's WITH_JS configure path no longer exists.
• Found mid-fix while seeding a clone for LOG-003; filed 2026-06-26.