Per adversarial review:
- Renamed `isHostArchInstructionSet` → `hasTargetStringMapping`. The
predicate just checks membership in `TARGET_SUBSTRINGS`; the old name
implied a "real CPU vs VM/IR/bytecode" distinction the implementation
doesn't make. Tightened the docstring: explicitly notes that real CPU
archs (`6502`, `mos6502`, `mrisc32`, `wdc65c816`) currently return
false because no production config exposes them via `-target=`, and
that the table needs extending if you add such a config. Pinned all
four with explicit assertions.
- Added `should have only valid InstructionSet values configured` test
in `properties-validation-tests.ts`. Catches typos like
`instructionSet=mos6052` that would silently slip through both
`findCompilersWithoutInstructionSet` (only checks presence) and the
consistency check (skips unknown values via the predicate).
- Documented the quote-stripping limitations in `extractTargetTokens`
with a `length >= 2` guard (avoids a bare `"` shrinking to `''`) and
three negative-case tests pinning down what actually happens for
leading-only quote, trailing-only quote, whitespace-inside-quotes,
and bare `"`. The behaviour is partly graceful (trailing quote ends
up in the extracted value, which is dirty but visible).
- Tightened the `lib/instructionsets.ts` predicate docstring to
"VM/bytecode" instead of "VM/IR" (LLVM IR is target-dependent so
classifying `llvm` as IR-style was overclaiming).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ae5926ea62