Beagle SCM

JS-017: no env / cwd — bind FILEGetEnv + FILEGetCwd

A JABC script can't read an environment variable or its working directory; both abc leaves already exist and just need marshalling. Method: Issues.

Input

Context

• No getenv / cwd reachable from JS (verified: grep getenv/getcwd in js/ → none).
• abc/FILE.h:518 FILEGetCwd(path8b) and :523 FILEGetEnv(name, u8cs) exist — bindable leaves, no new abc work.
• A utility can't find $HOME / config paths or report where it runs; require's .-relative resolve also wants a real cwd.

Goals

• io.getenv(name) and io.cwd() return the value / working directory.

Constraints

• Marshal over FILEGetEnv / FILEGetCwd; JS strings, no held refs.
• chdir has no abc wrapper today — out of scope (note as follow-up), don't block on it.

WIP

Design decisions

• (open) io.getenv(name) -> string|undefined, io.cwd() -> string.

TODOs

• Repro-first: a set env var reads back; io.cwd() matches a fixture.
• Bind the two FILE leaves.

Blockers

• None.

Outcome

• LANDED 8622368a (2026-06-21): io.cwd() over FILEGetCwd, io.getenv(name) over FILEGetEnv (unset → undefined; an empty-valued var is indistinguishable, documented), both in js/io.cpp + a JABCSliceStr helper. jabc_test.cpp rows getenv_set/getenv_unset/ cwd_matches red→green; ctest -R JABC 18/18, ASAN/LSan clean. Reconciled onto JS-015 via a clean be get --nosub weave-merge of INDEX.md/API.md; posted past the benign POSTCFLCT (--force).
• Follow-up: io.chdir needs a new FILEChdir abc leaf (none today).