c: add generator runtime crate

.clang-format

@@ -0,0 +1,11 @@
+BasedOnStyle: LLVM
+Language: Cpp
+ColumnLimit: 100
+UseTab: ForIndentation
+TabWidth: 8
+IndentWidth: 8
+BreakBeforeBraces: Linux
+IndentCaseLabels: false
+AllowShortFunctionsOnASingleLine: Empty
+SortIncludes: false
+PointerAlignment: Right

crates/sidex-gen-c/Cargo.toml

@@ -0,0 +1,15 @@
+[package]
+name = "sidex-gen-c"
+description = "Sidex code generation support for C."
+version = "0.1.0"
+edition.workspace = true
+authors.workspace = true
+license.workspace = true
+repository.workspace = true
+homepage.workspace = true
+
+[dependencies]
+sidex-gen.workspace = true
+
+[lints]
+workspace = true

crates/sidex-gen-c/README.md

@@ -0,0 +1,8 @@
+# Sidex: C Code Generation
+
+This crate is part of [Sidex](https://oss.silitics.com/sidex/). It contains
+the C code generation backend and the small portable C runtime used by
+generated C bindings.
+
+The backend is intentionally generic and independent of any product or RTOS.
+The C runtime is C99 and avoids dynamic allocation.

crates/sidex-gen-c/fuzz/README.md

@@ -0,0 +1,51 @@
+# Sidex C JSON Runtime Fuzzing
+
+This directory contains a libFuzzer target for the portable C JSON runtime.
+
+Build it with Clang:
+
+```sh
+mkdir -p target/sidex-gen-c-fuzz
+clang \
+  -std=c99 -g -O1 \
+  -fsanitize=fuzzer,address,undefined \
+  -I crates/sidex-gen-c/runtime \
+  crates/sidex-gen-c/runtime/sidex_types.c \
+  crates/sidex-gen-c/runtime/sidex_json_reader.c \
+  crates/sidex-gen-c/runtime/sidex_json_value.c \
+  crates/sidex-gen-c/runtime/sidex_json_writer.c \
+  crates/sidex-gen-c/fuzz/json_runtime_fuzzer.c \
+  -o target/sidex-gen-c-fuzz/json_runtime_fuzzer
+```
+
+Run it with the seed corpus:
+
+```sh
+rm -rf target/sidex-gen-c-fuzz/json_runtime_corpus
+mkdir -p target/sidex-gen-c-fuzz/json_runtime_corpus
+cp crates/sidex-gen-c/fuzz/corpus/json_runtime/* \
+  target/sidex-gen-c-fuzz/json_runtime_corpus/
+target/sidex-gen-c-fuzz/json_runtime_fuzzer \
+  target/sidex-gen-c-fuzz/json_runtime_corpus \
+  -max_len=16384 \
+  -rss_limit_mb=512
+```
+
+If Clang is not installed locally, the same commands can be run through Nix:
+
+```sh
+nix --extra-experimental-features 'nix-command flakes' shell nixpkgs#clang nixpkgs#clang-tools -c clang \
+  -std=c99 -g -O1 \
+  -fsanitize=fuzzer,address,undefined \
+  -I crates/sidex-gen-c/runtime \
+  crates/sidex-gen-c/runtime/sidex_types.c \
+  crates/sidex-gen-c/runtime/sidex_json_reader.c \
+  crates/sidex-gen-c/runtime/sidex_json_value.c \
+  crates/sidex-gen-c/runtime/sidex_json_writer.c \
+  crates/sidex-gen-c/fuzz/json_runtime_fuzzer.c \
+  -o target/sidex-gen-c-fuzz/json_runtime_fuzzer
+```
+
+The fuzzer treats malformed JSON, arena exhaustion, and output-buffer exhaustion
+as expected non-crashing outcomes. Successful decodes are encoded, parsed again,
+and encoded a second time; the two encoded byte strings must match.

crates/sidex-gen-c/fuzz/corpus/json_runtime/edge.json

@@ -0,0 +1 @@
+{"empty_object":{},"empty_array":[],"numbers":[0,1,-1,9007199254740991,1.0,1e+9]}

crates/sidex-gen-c/fuzz/corpus/json_runtime/nested.json

@@ -0,0 +1 @@
+{"items":[{"id":1,"tags":["a","b"]},{"id":2,"tags":[]}],"meta":{"ok":true}}

crates/sidex-gen-c/fuzz/corpus/json_runtime/null.json

@@ -0,0 +1 @@
+null

crates/sidex-gen-c/fuzz/corpus/json_runtime/scalars.json

@@ -0,0 +1 @@
+[true,false,null,0,-1,12.5e-3,"text","line\nbreak"]

crates/sidex-gen-c/fuzz/corpus/json_runtime/unicode.json

@@ -0,0 +1 @@
+{"raw":"ß🚀","escaped":"\u00df\ud83d\ude80","control":"a\u0000b"}

crates/sidex-gen-c/fuzz/json_runtime_fuzzer.c

@@ -0,0 +1,121 @@
+#include "sidex_json.h"
+
+#include <stddef.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+#ifndef SIDEX_FUZZ_MAX_INPUT
+#define SIDEX_FUZZ_MAX_INPUT 16384u
+#endif
+
+#ifndef SIDEX_FUZZ_ARENA_CAP
+#define SIDEX_FUZZ_ARENA_CAP 65536u
+#endif
+
+static void require(bool condition)
+{
+	if (!condition) {
+		abort();
+	}
+}
+
+static void exercise_skip_value(const uint8_t *data, size_t size)
+{
+	sidex_json_reader reader;
+	sidex_json_token_kind kind;
+	int rc;
+
+	sidex_json_reader_init(&reader, data, size);
+	rc = sidex_json_skip_value(&reader, SIDEX_JSON_DEFAULT_MAX_DEPTH);
+	if (rc == SIDEX_OK) {
+		(void)sidex_json_peek(&reader, &kind);
+	}
+}
+
+static void exercise_value_round_trip(const uint8_t *data, size_t size)
+{
+	uint8_t *arena_storage;
+	uint8_t *second_arena_storage;
+	uint8_t *output;
+	uint8_t *second_output;
+	size_t output_cap;
+	sidex_arena arena;
+	sidex_arena second_arena;
+	sidex_json_reader reader;
+	sidex_json_reader second_reader;
+	sidex_json_value value;
+	sidex_json_value second_value;
+	sidex_json_buffer buffer;
+	sidex_json_buffer second_buffer;
+	sidex_json_writer writer;
+	sidex_json_writer second_writer;
+	sidex_json_token_kind kind;
+	int rc;
+
+	arena_storage = (uint8_t *)malloc(SIDEX_FUZZ_ARENA_CAP);
+	second_arena_storage = (uint8_t *)malloc(SIDEX_FUZZ_ARENA_CAP);
+	output_cap = size * 8u + 256u;
+	output = (uint8_t *)malloc(output_cap);
+	second_output = (uint8_t *)malloc(output_cap);
+	if (arena_storage == NULL || second_arena_storage == NULL || output == NULL ||
+	    second_output == NULL) {
+		goto done;
+	}
+
+	arena = sidex_arena_from_parts(arena_storage, SIDEX_FUZZ_ARENA_CAP);
+	sidex_json_reader_init(&reader, data, size);
+	rc = sidex_json_read_value(&reader, &value, &arena, SIDEX_JSON_DEFAULT_MAX_DEPTH);
+	if (rc != SIDEX_OK) {
+		goto done;
+	}
+
+	rc = sidex_json_peek(&reader, &kind);
+	if (rc != SIDEX_OK) {
+		goto done;
+	}
+	if (kind != SIDEX_JSON_TOKEN_END) {
+		goto done;
+	}
+
+	sidex_json_buffer_init(&buffer, output, output_cap);
+	writer = sidex_json_writer_for_buffer(&buffer);
+	rc = sidex_json_write_value(&writer, &value, SIDEX_JSON_DEFAULT_MAX_DEPTH);
+	if (rc != SIDEX_OK) {
+		goto done;
+	}
+
+	second_arena = sidex_arena_from_parts(second_arena_storage, SIDEX_FUZZ_ARENA_CAP);
+	sidex_json_reader_init(&second_reader, output, buffer.len);
+	rc = sidex_json_read_value(&second_reader, &second_value, &second_arena,
+	                           SIDEX_JSON_DEFAULT_MAX_DEPTH);
+	require(rc == SIDEX_OK);
+	rc = sidex_json_peek(&second_reader, &kind);
+	require(rc == SIDEX_OK);
+	require(kind == SIDEX_JSON_TOKEN_END);
+
+	sidex_json_buffer_init(&second_buffer, second_output, output_cap);
+	second_writer = sidex_json_writer_for_buffer(&second_buffer);
+	rc = sidex_json_write_value(&second_writer, &second_value, SIDEX_JSON_DEFAULT_MAX_DEPTH);
+	require(rc == SIDEX_OK);
+	require(second_buffer.len == buffer.len);
+	require(memcmp(second_output, output, buffer.len) == 0);
+
+done:
+	free(second_output);
+	free(output);
+	free(second_arena_storage);
+	free(arena_storage);
+}
+
+int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size)
+{
+	if (size > SIDEX_FUZZ_MAX_INPUT) {
+		return 0;
+	}
+
+	(void)sidex_utf8_validate(data, size);
+	exercise_skip_value(data, size);
+	exercise_value_round_trip(data, size);
+	return 0;
+}

crates/sidex-gen-c/runtime/sidex_json.h

@@ -0,0 +1,10 @@
+#ifndef SIDEX_JSON_H
+#define SIDEX_JSON_H
+
+/** Umbrella include for the Sidex JSON reader and writer APIs. */
+
+#include "sidex_json_reader.h"
+#include "sidex_json_value.h"
+#include "sidex_json_writer.h"
+
+#endif