Add compilation corpus for fuzzing + improve CLI robustness of contrib/compile

contrib/compile.cc

@@ -1,6 +1,7 @@
 #include <sourcemeta/blaze/compiler.h>
 #include <sourcemeta/core/json.h>
 #include <sourcemeta/core/jsonpointer.h>
+#include <sourcemeta/core/jsonschema.h>
 #include <sourcemeta/core/options.h>
 
 #include <chrono>      // std::chrono
@@ -131,6 +132,14 @@ auto main(int argc, char **argv) noexcept -> int {
         sourcemeta::core::read_json(std::filesystem::path{positional.front()})};
     std::cerr << "Input: " << positional.front() << "\n";
 
+    // The compiler assumes valid schema input (object or boolean) and
+    // asserts otherwise. Guard here at the CLI boundary to ensure
+    // graceful failure instead of aborting on malformed inputs.
+    if (!sourcemeta::core::is_schema(document)) {
+      std::cerr << "error: the input is not a valid JSON Schema\n";
+      return EXIT_FAILURE;
+    }
+
     if (options.contains("path")) {
       const auto pointer{sourcemeta::core::to_pointer(
           std::string{options.at("path").front()})};
@@ -143,6 +152,12 @@ auto main(int argc, char **argv) noexcept -> int {
 
       auto extracted{*result};
       document = std::move(extracted);
+
+      if (!sourcemeta::core::is_schema(document)) {
+        std::cerr << "error: the value at the given path is not a valid "
+                     "JSON Schema\n";
+        return EXIT_FAILURE;
+      }
     }
 
     const auto compile_start{std::chrono::high_resolution_clock::now()};

contrib/run_corpus.sh

@@ -0,0 +1,67 @@
+#!/bin/sh
+# Helper script to run the compile binary against the corpus files.
+# Usage: ./contrib/run_corpus.sh <path-to-compile-binary>
+#
+# This is an optional utility for local testing and fuzzing preparation.
+# It is NOT part of the CI test suite.
+#
+# Expected behavior:
+#   - valid/ and stress/ files: should succeed (exit 0)
+#   - All other categories: should fail gracefully (exit non-zero), never crash
+
+set -e
+
+COMPILE="${1:?Usage: $0 <path-to-compile-binary>}"
+CORPUS_DIR="$(cd "$(dirname "$0")/../test/corpus/compile" && pwd)"
+
+if [ ! -x "$COMPILE" ]; then
+  echo "error: compile binary not found or not executable: $COMPILE" >&2
+  exit 1
+fi
+
+pass=0
+fail=0
+crash=0
+
+for category in valid invalid_json invalid_schema unknown_dialect edge_cases stress; do
+  dir="$CORPUS_DIR/$category"
+  [ -d "$dir" ] || continue
+  for f in "$dir"/*.json; do
+    [ -f "$f" ] || continue
+    name="${category}/$(basename "$f")"
+
+    rc=0
+    "$COMPILE" "$f" > /dev/null 2>&1 || rc=$?
+
+    if [ $rc -ge 128 ]; then
+      echo "CRASH  $name (signal $((rc - 128)))"
+      crash=$((crash + 1))
+    elif [ "$category" = "valid" ] || [ "$category" = "stress" ]; then
+      if [ $rc -eq 0 ]; then
+        echo "PASS   $name"
+        pass=$((pass + 1))
+      else
+        echo "FAIL   $name (expected success, got exit $rc)"
+        fail=$((fail + 1))
+      fi
+    else
+      if [ $rc -ne 0 ]; then
+        echo "PASS   $name (graceful error, exit $rc)"
+        pass=$((pass + 1))
+      else
+        echo "NOTE   $name (unexpected success, exit 0)"
+        pass=$((pass + 1))
+      fi
+    fi
+  done
+done
+
+echo ""
+echo "Results: $pass passed, $fail failed, $crash crashed"
+
+if [ $crash -gt 0 ]; then
+  echo "ERROR: $crash files caused a crash!" >&2
+  exit 1
+fi
+
+exit 0

test/corpus/compile/README.md

@@ -0,0 +1,56 @@
+# Compile Corpus
+
+Seed corpus for fuzzing the `contrib/compile` binary with
+[AFL++](https://aflplus.plus/) or similar tools. Each file is a small,
+purposeful input that exercises a specific compiler path.
+
+This corpus is designed for fuzzing and robustness testing, not as a
+formal conformance test suite.
+
+## Categories
+
+| Directory          | Purpose                                          |
+|--------------------|--------------------------------------------------|
+| `valid/`           | Correct schemas across supported dialects         |
+| `invalid_json/`    | Malformed JSON rejected by the parser             |
+| `invalid_schema/`  | Valid JSON that violates JSON Schema constraints  |
+| `unknown_dialect/` | Schemas with unsupported `$schema` values         |
+| `edge_cases/`      | Unusual but possibly valid inputs and boundaries  |
+| `stress/`          | Large schemas to test performance and scale       |
+
+## Known Issues (Discovered via Corpus)
+
+The following inputs trigger assertion failures (SIGABRT) inside the
+compiler. They are kept in the corpus intentionally to document the
+crashes and to serve as regression inputs once fixed.
+
+| File | Input | Issue |
+|------|-------|-------|
+| `invalid_schema/additional_properties_as_number.json` | `"additionalProperties": 42` | Asserts on non-boolean/non-schema value |
+| `invalid_schema/items_as_string.json` | `"items": "invalid"` | Asserts on non-schema items value |
+| `invalid_schema/max_length_negative.json` | `"maxLength": -1` | Asserts on negative string constraint |
+| `invalid_schema/multiple_of_negative.json` | `"multipleOf": -5` | Asserts on negative multipleOf |
+| `edge_cases/empty_keyword_values.json` | `"allOf": [], "anyOf": []` | Asserts on empty applicator arrays |
+
+Expected: graceful error (exception + exit 1). Actual: `assert()` failure.
+
+## Fuzzing with AFL++
+
+```sh
+afl-fuzz -i test/corpus/compile/ -o findings/ -- \
+  ./build/contrib/sourcemeta_blaze_contrib_compile @@
+```
+
+## Helper Script
+
+An optional helper script is available at `contrib/run_corpus.sh` to run
+the full corpus locally and detect crashes:
+
+```sh
+./contrib/run_corpus.sh ./build/contrib/sourcemeta_blaze_contrib_compile
+```
+
+## Adding New Files
+
+Each file should be **small** and have a **single clear purpose**. Name files
+descriptively (e.g., `type_as_integer.json`, `deeply_nested_allof.json`).

test/corpus/compile/edge_cases/array_root.json

@@ -0,0 +1 @@
+[]

test/corpus/compile/edge_cases/boolean_false.json

@@ -0,0 +1 @@
+false

test/corpus/compile/edge_cases/boolean_true.json

@@ -0,0 +1 @@
+true

test/corpus/compile/edge_cases/dangling_ref.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$ref": "#/$defs/nonexistent"
+}

test/corpus/compile/edge_cases/deeply_nested_allof.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "allOf": [{"allOf": [{"allOf": [{"allOf": [{"allOf": [{"allOf": [{"allOf": [{"allOf": [{"allOf": [{"allOf": [{"type": "string"}]}]}]}]}]}]}]}]}]}]
+}

test/corpus/compile/edge_cases/duplicate_keys.json

@@ -0,0 +1,8 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "properties": {
+    "name": { "type": "string" },
+    "name": { "type": "integer" }
+  }
+}

test/corpus/compile/edge_cases/empty_enum.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "enum": []
+}

test/corpus/compile/edge_cases/empty_keyword_values.json

@@ -0,0 +1,9 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "properties": {},
+  "required": [],
+  "allOf": [],
+  "anyOf": [],
+  "oneOf": []
+}

test/corpus/compile/edge_cases/empty_object.json

@@ -0,0 +1 @@
+{}

test/corpus/compile/edge_cases/empty_property_name.json

@@ -0,0 +1,8 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "properties": {
+    "": { "type": "string" }
+  },
+  "required": [""]
+}

test/corpus/compile/edge_cases/extreme_numeric_bounds.json

@@ -0,0 +1,6 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "number",
+  "minimum": 1e308,
+  "maximum": 1e308
+}

test/corpus/compile/edge_cases/invalid_regex_pattern.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "pattern": "[invalid(regex"
+}

test/corpus/compile/edge_cases/number_root.json

@@ -0,0 +1 @@
+42

test/corpus/compile/edge_cases/self_referencing.json

@@ -0,0 +1,7 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$ref": "#/$defs/self",
+  "$defs": {
+    "self": { "$ref": "#/$defs/self" }
+  }
+}

test/corpus/compile/edge_cases/unicode_keys.json

@@ -0,0 +1,9 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "properties": {
+    "héllo": { "type": "string" },
+    "日本語": { "type": "string" },
+    "🎉": { "type": "boolean" }
+  }
+}

test/corpus/compile/edge_cases/unknown_keywords.json

@@ -0,0 +1,7 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "unknown_keyword": true,
+  "x-custom-extension": { "foo": "bar" },
+  "notAKeyword": [1, 2, 3]
+}

test/corpus/compile/invalid_json/missing_closing_brace.json

@@ -0,0 +1 @@
+{

test/corpus/compile/invalid_json/missing_comma.json

@@ -0,0 +1,5 @@
+{
+  "type": "string",
+  "minLength": 1
+  "maxLength": 10
+}

test/corpus/compile/invalid_json/multiple_roots.json

@@ -0,0 +1 @@
+{"type": "string"} {"type": "number"}

test/corpus/compile/invalid_json/plain_text.json

@@ -0,0 +1 @@
+hello world this is not json

test/corpus/compile/invalid_json/trailing_comma.json

@@ -0,0 +1 @@
+{"type": "string",}

test/corpus/compile/invalid_json/truncated.json

@@ -0,0 +1 @@
+{"type": "stri

test/corpus/compile/invalid_json/unquoted_key.json

@@ -0,0 +1 @@
+{type: "string"}

test/corpus/compile/invalid_schema/additional_properties_as_number.json

@@ -0,0 +1,5 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": "object",
+  "additionalProperties": 42
+}

test/corpus/compile/invalid_schema/items_as_string.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "items": "invalid"
+}

test/corpus/compile/invalid_schema/max_length_negative.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "maxLength": -1
+}

test/corpus/compile/invalid_schema/multiple_of_negative.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "multipleOf": -5
+}

test/corpus/compile/invalid_schema/multiple_of_zero.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "multipleOf": 0
+}

test/corpus/compile/invalid_schema/ref_as_number.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$ref": 42
+}

test/corpus/compile/invalid_schema/type_as_integer.json

@@ -0,0 +1,4 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "type": 123
+}