go/conformance/run.sh

#!/usr/bin/env bash
# SPDX-License-Identifier: BSD-3-Clause-Eco
#
# The two-backend proof, end to end.
#
#   1. regenerate Go AND C++ from the same schemas with the same zapgen
#   2. build both probes
#   3. compare what each BUILT, byte for byte
#   4. read each side's bytes back with BOTH readers and compare the fields
#   5. compare what each READ from a corpus of real chain bytes
#
# A pass means the C++ backend writes the bytes the Go backend writes and
# reads the same fields out of bytes neither of them wrote.
#
# The C++ runtime comes from github.com/zap-proto/cpp, pinned below. Set
# ZAP_CPP to a checkout to use one you already have.
#
# The corpus is a vectors.tsv in node2's format. Pass a path, or set
# ZAP_CORPUS. Without one, only the build comparison runs, and the script
# says so rather than reporting a pass it did not earn.
set -euo pipefail

ZAP_CPP_REV=075e43b14878af486a98974759b57a5881df157a
here=$(cd "$(dirname "$0")" && pwd)
repo=$(cd "$here/.." && pwd)
work=${TMPDIR:-/tmp}/zapgen-conformance
corpus=${1:-${ZAP_CORPUS:-}}

mkdir -p "$work"

# ── the C++ runtime ────────────────────────────────────────────────────────
if [ -z "${ZAP_CPP:-}" ]; then
    ZAP_CPP=$work/zap-cpp
    if [ ! -d "$ZAP_CPP/.git" ]; then
        git clone -q https://github.com/zap-proto/cpp "$ZAP_CPP"
    fi
    git -C "$ZAP_CPP" fetch -q origin
    git -C "$ZAP_CPP" checkout -q $ZAP_CPP_REV
fi
echo "runtime: $ZAP_CPP @ $(git -C "$ZAP_CPP" rev-parse --short HEAD)"

# ── regenerate ─────────────────────────────────────────────────────────────
cd "$repo"
go build -o "$work/zapgen" ./cmd/zapgen

gen() { # gen SCHEMA GO_PKG_DIR
    "$work/zapgen" -single -out "conformance/gen/$2" "$1"
    "$work/zapgen" -lang cpp -single -out conformance/gen/hpp "$1"
}
gen conformance/schema/probe.zap probe
gen conformance/schema/pchain.zap pchain
gen cmd/zapgen/testdata/echo.zap echo
gen cmd/zapgen/testdata/basetx.zap xvm

# ── the per-struct headers compile too ─────────────────────────────────────
# -single is what the probes use; the default is one header per struct, with
# a sibling include per nested struct. Both have to build.
per=$work/per-struct
rm -rf "$per" && mkdir -p "$per"
"$work/zapgen" -lang cpp -out "$per" conformance/schema/probe.zap
"$work/zapgen" -lang cpp -out "$per" conformance/schema/pchain.zap
"$work/zapgen" -lang cpp -out "$per" cmd/zapgen/testdata/echo.zap
"$work/zapgen" -lang cpp -out "$per" cmd/zapgen/testdata/basetx.zap
{
    echo "// Generated by conformance/run.sh: include every per-struct header."
    for h in "$per"/*.hpp; do echo "#include \"$(basename "$h")\""; done
    echo "int main() { return 0; }"
} > "$per/all.cpp"
${CXX:-g++} -std=c++23 -O1 -Wall -Wextra -I"$ZAP_CPP/include" -I"$per" \
    -o "$work/per-struct-all" "$per/all.cpp" "$ZAP_CPP/src/rpc.cpp"
echo "per-struct headers: $(ls "$per"/*.hpp | wc -l) compiled together"

# ── build both probes ──────────────────────────────────────────────────────
go build -o "$work/probe-go" ./conformance/go
${CXX:-g++} -std=c++23 -O1 -Wall -Wextra \
    -I"$ZAP_CPP/include" -Iconformance/gen/hpp \
    -o "$work/probe-cpp" conformance/cpp/probe.cpp "$ZAP_CPP/src/rpc.cpp"

# ── what each BUILT ────────────────────────────────────────────────────────
"$work/probe-go" write > "$work/go.write"
"$work/probe-cpp" write > "$work/cpp.write"
if diff -u "$work/go.write" "$work/cpp.write"; then
    echo "write parity: $(wc -l < "$work/go.write") cases identical"
else
    echo "write parity: FAILED"
    exit 1
fi

# ── each reader over the other language's bytes ────────────────────────────
# Four runs, so every combination is covered: each reader over its own bytes
# and over the other's. All four must print the same fields.
"$work/probe-go" back "$work/go.write" > "$work/go-reads-go"
"$work/probe-cpp" back "$work/go.write" > "$work/cpp-reads-go"
"$work/probe-go" back "$work/cpp.write" > "$work/go-reads-cpp"
"$work/probe-cpp" back "$work/cpp.write" > "$work/cpp-reads-cpp"
if diff -u "$work/go-reads-go" "$work/cpp-reads-go" &&
    diff -u "$work/go-reads-go" "$work/go-reads-cpp" &&
    diff -u "$work/go-reads-go" "$work/cpp-reads-cpp"; then
    echo "read-back parity: $(wc -l < "$work/go-reads-go") cases, both readers, both writers"
else
    echo "read-back parity: FAILED"
    exit 1
fi

# ── what each READ ─────────────────────────────────────────────────────────
if [ -z "$corpus" ]; then
    echo "read parity: SKIPPED (no corpus; pass a vectors.tsv path or set ZAP_CORPUS)"
    exit 0
fi
"$work/probe-go" read "$corpus" > "$work/go.read"
"$work/probe-cpp" read "$corpus" > "$work/cpp.read"
if diff -u "$work/go.read" "$work/cpp.read"; then
    n=$(wc -l < "$work/go.read")
    ok=$(grep -c 'parse=ok' "$work/go.read" || true)
    echo "read parity: $n corpus vectors identical ($ok parsed, $((n - ok)) refused by both)"
else
    echo "read parity: FAILED"
    exit 1
fi