tests: Add -fshort-wchar tests

These tests are very tricky - in general -fshort-wchar is a dangerous
option because, since the standard library is not compiled with it, you
can't use any functions from either C or C++ standard library without
getting ABI mismatch.

The reason we want to use this is to do coverage testing on UTF32->UTF16
and UTF16->UTF32 conversion paths, that generally aren't hit on
gcc/clang. To do this, we carefully work around any internal calls to
wcslen/wcscmp that pugixml might be doing and don't use any wstring
functions.

Makefile

@@ -99,6 +99,10 @@ $(BUILD)/%.o: %
 	@mkdir -p $(dir $@)
 	$(CXX) $< $(CXXFLAGS) -c -MMD -MP -o $@
 
+$(BUILD)/tests/test_short_wchar.cpp.o: tests/test_short_wchar.cpp
+	@mkdir -p $(dir $@)
+	$(CXX) $< $(CXXFLAGS) -fshort-wchar -c -MMD -MP -o $@
+
 -include $(OBJECTS:.o=.d)
 
 .SECONDEXPANSION:

tests/test_short_wchar.cpp

@@ -0,0 +1,80 @@
+#if defined(__GNUC__) && __WCHAR_MAX__ == 0xffff
+#define PUGIXML_HEADER_ONLY
+#define PUGIXML_WCHAR_MODE
+#define PUGIXML_NO_STL
+#define pugi pugisw
+
+#include "test.hpp"
+
+#include <string.h>
+
+using namespace pugi;
+
+TEST(short_wchar_parse_utf32_be)
+{
+	xml_document doc;
+	CHECK(doc.load_buffer("\x00\x00\xfe\xff\x00\x00\x00t\x00\x00\x00o\x00\x00\x00s\x00\x00\x00t", 20, parse_fragment));
+
+	const char_t* text = doc.text().get();
+
+	CHECK(text[0] == 't');
+	CHECK(text[1] == 'o');
+	CHECK(text[2] == 's');
+	CHECK(text[3] == 't');
+	CHECK(text[4] == 0);
+}
+
+TEST(short_wchar_parse_utf32_le)
+{
+	xml_document doc;
+	CHECK(doc.load_buffer("\xff\xfe\x00\x00t\x00\x00\x00o\x00\x00\x00s\x00\x00\x00t\x00\x00\x00", 20, parse_fragment));
+
+	const char_t* text = doc.text().get();
+
+	CHECK(text[0] == 't');
+	CHECK(text[1] == 'o');
+	CHECK(text[2] == 's');
+	CHECK(text[3] == 't');
+	CHECK(text[4] == 0);
+}
+
+struct buffer_writer: xml_writer
+{
+	char buffer[128];
+	size_t offset;
+
+	buffer_writer(): offset(0)
+	{
+	}
+
+	virtual void write(const void* data, size_t size) PUGIXML_OVERRIDE
+	{
+		CHECK(offset + size <= sizeof(buffer));
+
+		memcpy(buffer + offset, data, size);
+		offset += size;
+	}
+};
+
+TEST(short_wchar_output_utf32_be)
+{
+	xml_document doc;
+	doc.append_child(node_doctype);
+
+	buffer_writer writer;
+	doc.print(writer, L"", format_raw, encoding_utf32_be);
+
+	CHECK(memcmp(writer.buffer, "\0\0\0<\0\0\0!\0\0\0D\0\0\0O\0\0\0C\0\0\0T\0\0\0Y\0\0\0P\0\0\0E\0\0\0>", 40) == 0);
+}
+
+TEST(short_wchar_output_utf32_le)
+{
+	xml_document doc;
+	doc.append_child(node_doctype);
+
+	buffer_writer writer;
+	doc.print(writer, L"", format_raw, encoding_utf32_le);
+
+	CHECK(memcmp(writer.buffer, "<\0\0\0!\0\0\0D\0\0\0O\0\0\0C\0\0\0T\0\0\0Y\0\0\0P\0\0\0E\0\0\0>\0\0\0", 40) == 0);
+}
+#endif