commit 9df676723aaea35b2c541a711b6219dc74a50c46
parent a5249ddd82cee9881f5db480ce7773c660453804
Author: Ryan Sepassi <rsepassi@gmail.com>
Date: Sun, 26 Apr 2026 02:39:12 -0700
make: wire cc-* suites into run-tests.sh + add cc-e2e
Six new suites plumbed through the existing Make + run-tests.sh
machinery, mirroring the m1pp / p1 / scheme1 pattern:
make test SUITE=cc-util 14 fixtures
make test SUITE=cc-lex 16 fixtures
make test SUITE=cc-pp 22 fixtures
make test SUITE=cc-cg 15 fixtures
make test SUITE=cc-parse 15 fixtures
make test SUITE=cc-e2e 1 fixture (00-return-argc -> ELF -> run)
Bare `make test` (no SUITE) now sweeps all nine suites.
Build artifacts:
build/$ARCH/cc/cc.scm — catm'd seven-file compiler source.
Per-arch by directory naming, identical
content (catm runs in the per-arch
container).
make cc — builds the artifact for $ARCH.
Suite-internal infrastructure files renamed with `_` prefix so
discover() ignores them automatically (matches the existing convention
that filenames starting with `_` are skipped):
tests/cc-lex/run-lex.scm -> _run-lex.scm
tests/cc-pp/run-pp.scm -> _run-pp.scm
tests/cc-parse/cg-trace.scm -> _cg-trace.scm
tests/cc-parse/mini-prelude.scm -> _mini-prelude.scm
tests/cc-lex/run.sh deleted (subsumed by run-tests.sh)
Two `.expected` files restored for cc-util's die tests (09, 10) and
two added for cc-pp's negative tests (#error, #include) so the
merged-stdout+stderr capture path matches the scheme1 suite's
convention (every fixture has a deterministic .expected).
Lessons-learned (recorded in run-tests.sh comments):
- `cmd >file 2>&1; rc=$?` swallows non-zero exit under busybox sh's
redirection semantics. All cc suite runners use the if/else pattern
that the existing scheme1 suite already uses.
- `expout=$(grep -v '^;;' file)` exits non-zero when zero matches,
killing the script. Guarded with `|| true`.
Diffstat:
11 files changed, 239 insertions(+), 111 deletions(-)
diff --git a/Makefile b/Makefile
@@ -16,6 +16,7 @@
# make pokem build pokem for ARCH
# make hello build hello via the bootstrap chain
# make scheme1 build the scheme1 interpreter for ARCH
+# make cc catm the cc compiler source for ARCH
# make run run hello in the container
# make test every suite, every arch
# make test SUITE=m1pp m1pp suite, every arch
@@ -61,7 +62,7 @@ PODMAN = podman run --rm --pull=never --platform $(PLATFORM_$(1)) \
# --- Targets --------------------------------------------------------------
-.PHONY: all m1pp pokem hello scheme1 run test image tools tables \
+.PHONY: all m1pp pokem hello scheme1 cc run test image tools tables \
tools-native cloc clean help
all: m1pp pokem
@@ -156,10 +157,18 @@ HELLO_BINS := $(foreach a,$(ALL_ARCHES),build/$(a)/hello)
SCHEME1_SRC := scheme1/scheme1.P1pp
SCHEME1_BINS := $(foreach a,$(ALL_ARCHES),build/$(a)/scheme1)
+# Catm'd cc compiler source. Per-arch only because catm runs in the
+# per-arch container; the resulting .scm is identical across arches but
+# we keep it under build/$arch/ for naming consistency.
+CC_SRCS := scheme1/prelude.scm cc/util.scm cc/data.scm cc/lex.scm \
+ cc/pp.scm cc/cg.scm cc/parse.scm cc/main.scm
+CC_BINS := $(foreach a,$(ALL_ARCHES),build/$(a)/cc/cc.scm)
+
m1pp: $(OUT_DIR)/m1pp
pokem: $(OUT_DIR)/pokem
hello: $(OUT_DIR)/hello
scheme1: $(OUT_DIR)/scheme1
+cc: $(OUT_DIR)/cc/cc.scm
# Per-arch deps for .P1/.M1 builds (raw M1, no macro expansion).
P1_BUILD_DEPS = scripts/lint.sh scripts/boot-build-p1.sh \
@@ -187,6 +196,12 @@ $(HELLO_BINS): build/%/hello: $(HELLO_SRC) $(P1_BUILD_DEPS)
$(SCHEME1_BINS): build/%/scheme1: $(SCHEME1_SRC) $(P1PP_BUILD_DEPS)
$(call PODMAN,$*) sh scripts/boot-build-p1pp.sh $(SCHEME1_SRC) $@
+# cc.scm: catm prelude + the six cc/*.scm files into one source the
+# scheme1 interpreter can run. Catm runs inside the per-arch container.
+$(CC_BINS): build/%/cc/cc.scm: $(CC_SRCS) build/%/.image build/%/tools/M0
+ mkdir -p $(@D)
+ $(call PODMAN,$*) build/$*/tools/catm $@ $(CC_SRCS)
+
run: $(OUT_DIR)/hello $(IMAGE_STAMP)
$(call PODMAN,$(ARCH)) ./$(OUT_DIR)/hello
@@ -237,11 +252,27 @@ TEST_P1_DEPS := $(foreach a,$(TEST_ARCHES), \
TEST_SCHEME1_DEPS := $(foreach a,$(TEST_ARCHES), \
build/$(a)/.image build/$(a)/tools/M0 build/$(a)/m1pp build/$(a)/scheme1)
+# cc-* unit suites (lex / pp / cg / parse / util) just need scheme1 +
+# the catm'd cc compiler source as artifacts.
+TEST_CC_UNIT_DEPS := $(foreach a,$(TEST_ARCHES), \
+ build/$(a)/.image build/$(a)/tools/M0 build/$(a)/m1pp build/$(a)/scheme1 \
+ build/$(a)/cc/cc.scm)
+
+# cc-e2e additionally needs the P1pp toolchain to assemble cc-emitted
+# P1pp into native ELF.
+TEST_CC_E2E_DEPS := $(TEST_CC_UNIT_DEPS)
+
test:
ifeq ($(SUITE),)
@$(MAKE) --no-print-directory test SUITE=m1pp
@$(MAKE) --no-print-directory test SUITE=p1
@$(MAKE) --no-print-directory test SUITE=scheme1
+ @$(MAKE) --no-print-directory test SUITE=cc-util
+ @$(MAKE) --no-print-directory test SUITE=cc-lex
+ @$(MAKE) --no-print-directory test SUITE=cc-pp
+ @$(MAKE) --no-print-directory test SUITE=cc-cg
+ @$(MAKE) --no-print-directory test SUITE=cc-parse
+ @$(MAKE) --no-print-directory test SUITE=cc-e2e
else ifeq ($(SUITE),m1pp)
@$(MAKE) --no-print-directory $(TEST_M1PP_DEPS)
sh scripts/run-tests.sh --suite=m1pp $(if $(ARCH_FILTER),--arch=$(ARCH_FILTER))
@@ -251,6 +282,12 @@ else ifeq ($(SUITE),p1)
else ifeq ($(SUITE),scheme1)
@$(MAKE) --no-print-directory $(TEST_SCHEME1_DEPS)
sh scripts/run-tests.sh --suite=scheme1 $(if $(ARCH_FILTER),--arch=$(ARCH_FILTER))
+else ifeq ($(filter $(SUITE),cc-util cc-lex cc-pp cc-cg cc-parse),$(SUITE))
+ @$(MAKE) --no-print-directory $(TEST_CC_UNIT_DEPS)
+ sh scripts/run-tests.sh --suite=$(SUITE) $(if $(ARCH_FILTER),--arch=$(ARCH_FILTER))
+else ifeq ($(SUITE),cc-e2e)
+ @$(MAKE) --no-print-directory $(TEST_CC_E2E_DEPS)
+ sh scripts/run-tests.sh --suite=cc-e2e $(if $(ARCH_FILTER),--arch=$(ARCH_FILTER))
else
- @echo "unknown SUITE='$(SUITE)' (expected m1pp | p1 | scheme1)" >&2; exit 2
+ @echo "unknown SUITE='$(SUITE)' (m1pp | p1 | scheme1 | cc-util | cc-lex | cc-pp | cc-cg | cc-parse | cc-e2e)" >&2; exit 2
endif
diff --git a/scripts/run-tests.sh b/scripts/run-tests.sh
@@ -47,8 +47,8 @@ while [ "$#" -gt 0 ]; do
done
case "$SUITE" in
- m1pp|p1|scheme1) ;;
- "") echo "$0: --suite required (m1pp | p1 | scheme1)" >&2; exit 2 ;;
+ m1pp|p1|scheme1|cc-util|cc-lex|cc-pp|cc-cg|cc-parse|cc-e2e) ;;
+ "") echo "$0: --suite required (m1pp | p1 | scheme1 | cc-util | cc-lex | cc-pp | cc-cg | cc-parse | cc-e2e)" >&2; exit 2 ;;
*) echo "$0: unknown suite '$SUITE'" >&2; exit 2 ;;
esac
@@ -280,10 +280,201 @@ run_scheme1_suite() {
done
}
+## --- cc-* suites --------------------------------------------------------
+##
+## Six suites, all sharing the same shape: catm a list of layer files
+## into a per-fixture combined .scm, run scheme1 on the result, diff
+## merged stdout+stderr against an expected-output file, diff exit
+## status against .expected-exit (default 0).
+##
+## Caller (Make) ensures build/<arch>/scheme1 and build/<arch>/cc/cc.scm
+## already exist. The cc-e2e suite additionally requires the P1pp
+## toolchain (m1pp + tools), which the Make deps already cover.
+##
+## Filenames starting with `_` are skipped by `discover()` — that's how
+## suite-internal driver and mock files (_run-lex.scm, _cg-trace.scm,
+## _mini-prelude.scm, _run-pp.scm) avoid being picked up as fixtures.
+
+# _cc_check <label> <expected-stdout> <expected-exit> <actual-stdout> <actual-exit>
+_cc_check() {
+ lbl=$1; exp_out=$2; exp_exit=$3; act_out=$4; act_exit=$5
+ if [ "$act_out" = "$exp_out" ] && [ "$act_exit" = "$exp_exit" ]; then
+ report "$lbl" PASS
+ else
+ report "$lbl" FAIL
+ if [ "$act_out" != "$exp_out" ]; then show_diff "$exp_out" "$act_out"; fi
+ if [ "$act_exit" != "$exp_exit" ]; then
+ echo " exit: expected $exp_exit, got $act_exit"
+ fi
+ fi
+}
+
+# _cc_arches: aarch64 amd64 riscv64 by default; honors --arch=...
+_cc_arches() {
+ if [ -z "$ARCH" ]; then echo "aarch64 amd64 riscv64"
+ else echo "$ARCH"; fi
+}
+
+# _cc_unit_suite <suite-name> <expected-ext> <layer-list>
+# Generic unit-suite runner for cc-util / cc-pp / cc-cg / cc-parse.
+# Each fixture is `tests/<suite>/<name>.scm`; expected stdout in
+# `tests/<suite>/<name>.<expected-ext>` (default empty); exit in
+# `tests/<suite>/<name>.expected-exit` (default 0).
+_cc_unit_suite() {
+ suite=$1; ext=$2; layers=$3
+ [ -n "$NAMES" ] || NAMES=$(discover tests/$suite scm)
+ for arch in $(_cc_arches); do
+ for name in $NAMES; do
+ fixture=tests/$suite/$name.scm
+ [ -e "$fixture" ] || { echo " SKIP $name (no .scm)"; continue; }
+ if [ -e "tests/$suite/$name.$ext" ]; then
+ expout=$(cat "tests/$suite/$name.$ext")
+ else
+ expout=
+ fi
+ if [ -e "tests/$suite/$name.expected-exit" ]; then
+ expexit=$(cat "tests/$suite/$name.expected-exit")
+ else
+ expexit=0
+ fi
+ tmp=$(mktemp)
+ if run_in_container "$arch" sh -c "
+ build/$arch/tools/catm /tmp/cc-test.scm $layers $fixture
+ exec build/$arch/scheme1 /tmp/cc-test.scm
+ " >"$tmp" 2>&1; then
+ act_exit=0
+ else
+ act_exit=$?
+ fi
+ act_out=$(cat "$tmp"); rm -f "$tmp"
+ _cc_check "[$arch] $suite/$name" "$expout" "$expexit" "$act_out" "$act_exit"
+ done
+ done
+}
+
+# cc-util: scheme1 prelude + util only.
+run_cc_util_suite() {
+ _cc_unit_suite cc-util expected "scheme1/prelude.scm cc/util.scm"
+}
+
+# cc-pp: prelude + util + data + pp.
+run_cc_pp_suite() {
+ _cc_unit_suite cc-pp expected \
+ "scheme1/prelude.scm cc/util.scm cc/data.scm cc/pp.scm"
+}
+
+# cc-cg: prelude + util + data + cg.
+run_cc_cg_suite() {
+ _cc_unit_suite cc-cg expected \
+ "scheme1/prelude.scm cc/util.scm cc/data.scm cc/cg.scm"
+}
+
+# cc-parse: mini-prelude + util + data + cg-trace mock + parse.
+# (Full prelude+everything would push past scheme1's source-buffer cap
+# — see CC-INTERNALS.md and the project memory note.)
+run_cc_parse_suite() {
+ _cc_unit_suite cc-parse expected-trace \
+ "tests/cc-parse/_mini-prelude.scm cc/util.scm cc/data.scm tests/cc-parse/_cg-trace.scm cc/parse.scm"
+}
+
+# cc-lex: needs the _run-lex.scm driver + the .c fixture as scheme1 argv.
+# Expected file is .expected-toks; `;;` lines are stripped (negative
+# fixtures use them for human notes).
+run_cc_lex_suite() {
+ [ -n "$NAMES" ] || NAMES=$(discover tests/cc-lex c)
+ layers="scheme1/prelude.scm cc/util.scm cc/data.scm cc/lex.scm tests/cc-lex/_run-lex.scm"
+ for arch in $(_cc_arches); do
+ for name in $NAMES; do
+ fixture=tests/cc-lex/$name.c
+ [ -e "$fixture" ] || { echo " SKIP $name (no .c)"; continue; }
+ if [ -e "tests/cc-lex/$name.expected-toks" ]; then
+ # `|| true` because grep returns 1 when zero lines match
+ # (negative-test fixtures have only ;; comments).
+ expout=$(grep -v '^;;' "tests/cc-lex/$name.expected-toks" || true)
+ else
+ expout=
+ fi
+ if [ -e "tests/cc-lex/$name.expected-exit" ]; then
+ expexit=$(cat "tests/cc-lex/$name.expected-exit")
+ else
+ expexit=0
+ fi
+ tmp=$(mktemp)
+ # Negative fixtures (expected-exit != 0) write diagnostics to
+ # stderr; we don't compare their output, only exit code.
+ # The merged stdout+stderr capture is uniform either way.
+ if run_in_container "$arch" sh -c "
+ build/$arch/tools/catm /tmp/cc-test.scm $layers
+ exec build/$arch/scheme1 /tmp/cc-test.scm $fixture
+ " >"$tmp" 2>&1; then
+ act_exit=0
+ else
+ act_exit=$?
+ fi
+ if [ "$expexit" != "0" ]; then
+ act_out=
+ else
+ act_out=$(cat "$tmp")
+ fi
+ rm -f "$tmp"
+ _cc_check "[$arch] cc-lex/$name" "$expout" "$expexit" "$act_out" "$act_exit"
+ done
+ done
+}
+
+# cc-e2e: compile a .c through cc, assemble to ELF, run.
+# Fixture: <name>.c; expected stdout in <name>.expected (default empty);
+# exit in <name>.expected-exit (default 0).
+run_cc_e2e_suite() {
+ [ -n "$NAMES" ] || NAMES=$(discover tests/cc-e2e c)
+ for arch in $(_cc_arches); do
+ for name in $NAMES; do
+ src=tests/cc-e2e/$name.c
+ [ -e "$src" ] || { echo " SKIP $name (no .c)"; continue; }
+ expout=$([ -e tests/cc-e2e/$name.expected ] \
+ && cat tests/cc-e2e/$name.expected || echo "")
+ expexit=$([ -e tests/cc-e2e/$name.expected-exit ] \
+ && cat tests/cc-e2e/$name.expected-exit || echo 0)
+ outdir=build/$arch/cc-e2e/$name
+ p1pp=$outdir/$name.P1pp
+ elf=$outdir/$name
+ mkdir -p "$outdir"
+ if ! run_in_container "$arch" sh -c \
+ "build/$arch/scheme1 build/$arch/cc/cc.scm $src $p1pp" \
+ >/dev/null 2>&1; then
+ report "[$arch] cc-e2e/$name" FAIL
+ echo " cc compile failed"
+ continue
+ fi
+ if ! run_in_container "$arch" sh scripts/boot-build-p1pp.sh \
+ "$p1pp" "$elf" >/dev/null 2>&1; then
+ report "[$arch] cc-e2e/$name" FAIL
+ run_in_container "$arch" sh scripts/boot-build-p1pp.sh \
+ "$p1pp" "$elf" 2>&1 | sed 's/^/ /' >&2 || true
+ continue
+ fi
+ tmp=$(mktemp)
+ if run_in_container "$arch" "./$elf" >"$tmp" 2>&1; then
+ act_exit=0
+ else
+ act_exit=$?
+ fi
+ act_out=$(cat "$tmp"); rm -f "$tmp"
+ _cc_check "[$arch] cc-e2e/$name" "$expout" "$expexit" "$act_out" "$act_exit"
+ done
+ done
+}
+
case "$SUITE" in
- m1pp) run_m1pp_suite ;;
- p1) run_p1_suite ;;
- scheme1) run_scheme1_suite ;;
+ m1pp) run_m1pp_suite ;;
+ p1) run_p1_suite ;;
+ scheme1) run_scheme1_suite ;;
+ cc-util) run_cc_util_suite ;;
+ cc-lex) run_cc_lex_suite ;;
+ cc-pp) run_cc_pp_suite ;;
+ cc-cg) run_cc_cg_suite ;;
+ cc-parse) run_cc_parse_suite ;;
+ cc-e2e) run_cc_e2e_suite ;;
esac
echo "$PASS passed, $FAIL failed"
diff --git a/tests/cc-lex/run-lex.scm b/tests/cc-lex/_run-lex.scm
diff --git a/tests/cc-lex/run.sh b/tests/cc-lex/run.sh
@@ -1,104 +0,0 @@
-#!/bin/sh
-## tests/cc-lex/run.sh — fixture-by-fixture lex test loop.
-##
-## Usage:
-## tests/cc-lex/run.sh # run all fixtures
-## tests/cc-lex/run.sh 02 05 # run only matching prefixes
-##
-## Per-fixture protocol:
-## - <NN>-<name>.c is the input.
-## - <NN>-<name>.expected-toks is the golden token serialization
-## (one tok per line, per CC-CONTRACTS §2.1). Lines beginning
-## with `;;` are ignored, allowing negative tests to carry a
-## human-readable note in their .expected-toks file.
-## - <NN>-<name>.expected-exit overrides the expected exit status
-## (default: 0).
-##
-## Each fixture is run inside boot2-busybox:aarch64 by catm'ing
-## prelude+lex+data+util+driver, then invoking scheme1 on the combined
-## file with the .c fixture as argv[2]. Stdout is diffed against the
-## (filtered) .expected-toks; exit status is diffed against
-## .expected-exit. stderr is suppressed (negative tests only check
-## exit status).
-
-set -eu
-REPO=$(cd "$(dirname "$0")/../.." && pwd)
-cd "$REPO"
-
-ARCH=${ARCH:-aarch64}
-case "$ARCH" in
- aarch64) PLATFORM=linux/arm64 ;;
- amd64) PLATFORM=linux/amd64 ;;
- riscv64) PLATFORM=linux/riscv64 ;;
- *) echo "unknown ARCH=$ARCH" >&2; exit 2 ;;
-esac
-
-PASS=0
-FAIL=0
-
-discover() {
- ls tests/cc-lex \
- | sed -n 's/^\([^_][^.]*\)\.c$/\1/p' \
- | sort -u
-}
-
-NAMES="$*"
-[ -n "$NAMES" ] || NAMES=$(discover)
-
-# Strip ;; comment lines from a file; emit the rest. Used to filter
-# expected-toks so negative tests can carry a human note.
-strip_comments() {
- sed -n '/^;;/d; p' "$1"
-}
-
-for prefix in $NAMES; do
- for c in tests/cc-lex/${prefix}*.c; do
- [ -e "$c" ] || continue
- name=$(basename "$c" .c)
- expected_toks=tests/cc-lex/$name.expected-toks
- expected_exit_file=tests/cc-lex/$name.expected-exit
-
- if [ ! -e "$expected_toks" ]; then
- echo " SKIP $name (no .expected-toks)"
- continue
- fi
- expected=$(strip_comments "$expected_toks")
- if [ -e "$expected_exit_file" ]; then
- expected_exit=$(cat "$expected_exit_file" | tr -d '\n')
- else
- expected_exit=0
- fi
-
- # Build the combined source and run.
- actual=$(podman run --rm --pull=never --platform "$PLATFORM" \
- --tmpfs /tmp:size=512M -v "$REPO":/work -w /work \
- "boot2-busybox:$ARCH" \
- sh -c "build/$ARCH/tools/catm /tmp/c.scm \
- scheme1/prelude.scm cc/util.scm cc/data.scm \
- cc/lex.scm tests/cc-lex/run-lex.scm; \
- build/$ARCH/scheme1 /tmp/c.scm tests/cc-lex/$name.c" \
- 2>/dev/null) || actual_exit=$?
- actual_exit=${actual_exit:-0}
-
- if [ "$actual" = "$expected" ] && [ "$actual_exit" = "$expected_exit" ]; then
- PASS=$((PASS + 1))
- echo " PASS $name"
- else
- FAIL=$((FAIL + 1))
- echo " FAIL $name"
- if [ "$actual" != "$expected" ]; then
- echo " --- expected ---"
- printf '%s\n' "$expected" | sed 's/^/ /'
- echo " --- actual ---"
- printf '%s\n' "$actual" | sed 's/^/ /'
- fi
- if [ "$actual_exit" != "$expected_exit" ]; then
- echo " exit: expected $expected_exit, got $actual_exit"
- fi
- fi
- unset actual_exit
- done
-done
-
-echo "$PASS passed, $FAIL failed"
-[ "$FAIL" -eq 0 ]
diff --git a/tests/cc-parse/cg-trace.scm b/tests/cc-parse/_cg-trace.scm
diff --git a/tests/cc-parse/mini-prelude.scm b/tests/cc-parse/_mini-prelude.scm
diff --git a/tests/cc-pp/16-error.expected b/tests/cc-pp/16-error.expected
@@ -0,0 +1 @@
+t.c:1:2: error: #error: boom
diff --git a/tests/cc-pp/17-include-rejected.expected b/tests/cc-pp/17-include-rejected.expected
@@ -0,0 +1 @@
+t.c:1:10: error: #include: file inclusion is handled upstream by pre-flatten
diff --git a/tests/cc-pp/run-pp.scm b/tests/cc-pp/_run-pp.scm
diff --git a/tests/cc-util/09-die-noloc.expected b/tests/cc-util/09-die-noloc.expected
@@ -0,0 +1 @@
+error: bad token: unexpected 42
diff --git a/tests/cc-util/10-die-no-irritants.expected b/tests/cc-util/10-die-no-irritants.expected
@@ -0,0 +1 @@
+error: something went wrong
