Fuzz the streaming input adapters against the contiguous one

The fast paths for contiguous input (bulk string scanning, multi-byte
loads, lazy position tracking) are taken only when parsing from a
pointer range; streams and non-contiguous iterators still read byte by
byte. The fuzzer drivers parsed only contiguous input, so a difference
between the two would ship unnoticed.

check_input_adapters() in fuzzer_common.hpp reads the input through a
std::istringstream, a std::list iterator range, and (on POSIX) a FILE*
from fmemopen, and asserts the same value, or the same exception id,
message and byte position, as for the pointer range. All seven drivers
call it; the JSON driver checks parse and accept, each with and without
ignore_comments. It replaces the stream check of the BON8 driver.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

tests/src/fuzzer_common.hpp

@@ -11,6 +11,14 @@
 #pragma once
 
 #include <cassert> // assert
+#include <cstddef> // size_t
+#include <cstdint> // uint8_t
+#include <cstdio> // FILE, fmemopen, fclose
+#include <list> // list
+#include <sstream> // istringstream
+#include <string> // string
+#include <utility> // forward
+#include <vector> // vector
 #include <nlohmann/json.hpp>
 
 using nlohmann::json; // NOLINT(google-global-names-in-headers): shared by all fuzzer drivers
@@ -49,3 +57,89 @@ json parse_without_exceptions(Parse parse, bool& threw)
     }
     return {};
 }
+
+// the outcome of reading an input with exceptions: the serialization of the
+// value, or the id, message, and (for parse errors) byte of the exception
+struct read_outcome
+{
+    std::string value{};
+    int id = 0;
+    std::string what{};
+    std::size_t byte = 0;
+
+    bool operator==(const read_outcome& rhs) const
+    {
+        return value == rhs.value && id == rhs.id && what == rhs.what && byte == rhs.byte;
+    }
+};
+
+inline std::string outcome_value(const json& j)
+{
+    return j.dump(-1, ' ', false, json::error_handler_t::keep);
+}
+
+inline std::string outcome_value(bool b)
+{
+    return b ? "true" : "false";
+}
+
+template<typename Read, typename... Args>
+read_outcome read_with(const Read& read, Args&& ... args)
+{
+    read_outcome result;
+    try
+    {
+        result.value = outcome_value(read(std::forward<Args>(args)...));
+    }
+    catch (const json::parse_error& e)
+    {
+        result.id = e.id;
+        result.what = e.what();
+        result.byte = e.byte;
+    }
+    catch (const json::exception& e)
+    {
+        result.id = e.id;
+        result.what = e.what();
+    }
+    return result;
+}
+
+// The contiguous input (a pointer range) takes the bulk fast paths; the other
+// input adapters read byte by byte. Reads @a data through each of them and
+// checks that all give the same value, or the same exception with the same
+// id, message, and byte position.
+//
+// @a read is called with either an iterator range or a single input (stream
+// or FILE*) and returns the parsed value (or the result of accept()).
+template<typename Read>
+void check_input_adapters(const std::uint8_t* data, std::size_t size, const Read& read)
+{
+    const read_outcome expected = read_with(read, data, data + size);
+
+    // stream adapter
+    {
+        std::istringstream stream(std::string(reinterpret_cast<const char*>(data), size));
+        assert(read_with(read, stream) == expected);
+    }
+
+    // iterator adapter on a non-contiguous range
+    {
+        const std::list<std::uint8_t> list(data, data + size);
+        assert(read_with(read, list.begin(), list.end()) == expected);
+    }
+
+#if defined(__unix__) || defined(__APPLE__)
+    // FILE* adapter; fmemopen may refuse an empty buffer
+    {
+        std::vector<std::uint8_t> buffer(data, data + size);
+        std::FILE* file = buffer.empty() ? nullptr : fmemopen(buffer.data(), buffer.size(), "rb");
+        if (file != nullptr)
+        {
+            const read_outcome actual = read_with(read, file);
+            std::fclose(file);
+            assert(actual == expected);
+        }
+    }
+#endif
+}