SPARK_BC makefile targets

End-to-end gates and helpers for the Spark bytecode factory. Copy-paste from the repo root after make spark-bootstrap / make spark as needed. Spark / SparkLang only.

ISA: SPARK_BC.md. Story: SPARK_BC Builder.

Core gates

Target	Purpose
`make test-sparkbc`	`run_sparkbc.sh` + `tools/spark-bc-dump/test_dump.py` (compile, dump, serve, SGD asserts)
`make test-sparkbc-e2e`	Focused TRAIN→STEP→ARTIFACT (`tools/spark-bc-dump/run_e2e_gate.sh`)
`make sparkbc-e2e`	Alias for `test-sparkbc-e2e`
`make test-decompile-compete`	Richer dump symbols/xrefs + analysis project units
`make decompile-roundtrip`	compile→dump→recompile sha256 on published fixtures
`make decompile-bench`	SPARK_BC metrics + `website/data/decompile-scoreboard.json` (+ `elf_local_probe`)
`make test-spark-lsp`	Editor LSP static diagnostics gate
`make spark-sgd-proof`	Multi-pass CPU SGD → `checkpoint.json` loss drop → `make spark-eval WEIGHTS=…`
`make spark-sgd-proof-scale`	Local opt-in larger JSONL + dim/n_layer (scale fixtures; not default CI)
`make spark-coder-train`	Owned TinyCoder tiny (CI/default)
`make spark-coder-train-large`	Opt-in large coder (dim 64 / n_layer 4); not GHA default
`make weight-gallery`	Catalog + emit scale/large samples + website catalog JSON
`make test-weights-play`	Weight gallery unit + CLI play/diff/stats
`make weight-gallery-xl`	Opt-in XL emit
`make spark-eval`	Frozen copy/recall + next-token probes; exit 0 = harness ran (measurement only)
`make test-spark-eval`	Unit gate for eval harness
`make docs-html` / `make docs-check`	Regen `website/docs/*` + nav link check
`make spark-serve-api` / `make test-serve-api`	serve API path HTTP/stdio predict + embeddings
`make helpers` / `make tools-test`	Helper CLIs + kit/shadow/analyze smoke
`make test-spark-ask`	Voice/text ask loop (dump facts + dry TTS)
`make test-spark-analyze`	Project-loop analyze folder gate
`make sdk-pack` / `make dist`	Stage `dist/spark-sdk/` (helpers overlay when present)
`make voice-easy` / `make test-voice-easy`	Owned voice STT/TTS heads — tiny dry (CI)
`make voice-easy-large`	Opt-in large voice-easy (not default CI)

make test-sparkbc
        make sparkbc-e2e
        make spark-sgd-proof
        make spark-eval
        make docs-check
        make tools-test
        make sdk-pack
        make test-voice-easy

Build / tools

Target	Purpose
`make spark-bootstrap` / `make sparkc`	C bootstrap VM + `--compile` / `--run-bc`
`make spark`	GAS ELF (`./spark`)
`make spark-serve`	Install `./spark-serve` companion script
`make spark-bc`	Product wrapper script (`scripts/spark-bc`)
`make spark-bc-emit` / `make test-bc-emit`	SPARK_BC → sasm emit probe + pack hello
`make sparkasm` / `make test-sparkasm`	Peer assembler
`make test-sparkasm-control`	`control.sparkasm` shape check
`make test-model-lab`	`examples/model_lab.spark` dry + expects
`make test-bootstrap`	Bootstrap VM suite
`make test-bpe-seed`	Tokenizer BPE seed (when present)

Helpers / shadows

See TOOLS_HELPERS.md.

Path	Role
`helpers/spark-run`	compile → run-bc → dump
`helpers/spark-analyze`	project loop → `out/analyze/<name>/` (optional `--serve`/`--ask`)
`helpers/spark-train-proof`	wrap `spark-sgd-proof` (`SCALE=1` → scale)
`helpers/spark-check-env`	env / fixture / import check
`helpers/spark-bc-pp` / `spark-bc-diff`	pretty-print / diff `.sparkbc`
`helpers/spark-shadow`	shadow copy, `build/shadow/`, hash verify
`./spark-ask` / `./spark-speak-ask`	voice/text ask over dump / analysis dir
`tools/spark_kit/`	hexdump, opcode sheet, fixture lint, vocab inspect

Docs / site regen

make docs-html
        make docs-check
        # Mirror CHANGELOG.md → website/CHANGELOG.html for Pages when cutting

Release + Pages: CI_PAGES.md · RELEASE.md.

Scripts (not always phony targets)

Path	Role
`scripts/sparkbc-e2e`	Wrapper around e2e gate
`scripts/spark-bc`	Phase-6 compile→bc_vm product wrapper
`bootstrap/tests/run_sparkbc.sh`	Oracle compare for `test-sparkbc`
`tools/spark-bc-dump/apply_step.py`	CPU SGD helper used by `spark-sgd-proof`
`tools/spark-eval/run.py`	Eval harness
`tools/package_helpers_k.sh`	`dist/spark-sdk/` helpers stager

CI

GitHub Actions workflow .github/workflows/sparkbc.yml runs the SPARK_BC gates on PRs (includes make tools-test + make sdk-pack). Prefer green test-sparkbc + sparkbc-e2e before claiming factory docs match tip.

Never

Claim spark-eval scores as a competitive AI win.
Treat empty / failing gates as soft success.