Cpp contains string literals that are not identical in grammar

[kaby76]

Using Tsql, and Version 4.12.0 of the Antlr tool, generate a parser and lexer for both CSharp and Cpp targets. Then, look at the generated TsqlParser files.

In a generated table of Cpp target TsqlParser.cpp auto staticData = std::make_unique<TSqlParserStaticData>( and that in C# PlsqlParser.cs _LiteralNames, the very last entry in the string table "literalNames" for C# is '?', but it is '\\u003F' for the C++ code. U+003F is the Unicode for the ?-character. There is no string literal '\\u003F' in the lexer or parser grammar, just '?', here.

C#:

private static readonly string[] _LiteralNames = {
	null, "'ADD'", "'ALL'", "'ALTER'", "'AND'", "'ANY'", "'AS'", "'ASC'", 
	...
	"'&'", "'^'", "'?'"
};

C++:

auto staticData = std::make_unique<TSqlParserStaticData>(
	std::vector<std::string>{
		"tsql_file", "batch", "batch_level_statement", "sql_clauses", "dml_clause", 
		...
		},
	std::vector<std::string>{
		"", "'ADD'", "'ALL'", "'ALTER'", "'AND'", "'ANY'", "'AS'", "'ASC'", 
		...
		"'/'", "'%'", "'+'", "'-'", "'~'", "'|'", "'&'", "'^'", "'\\u003F'"
		},

The strings literals should be identical across all targets, and identical to the string literal in the grammar. This is necessary if error messages that occur of a parse are to remain identical. In my opinion, the generated C++ code is wrong.

Does this difference have something to do with something with fragment rules?