c++ regex_match native Ubuntu vs WSL

[T4lus]

I have some strange behavior while using both Ubuntu and Ubuntu on Win 10 (WSL). Both versions are 18.04 and GCC, and I'm compiling with these flags: -std=c++11 -g.

My problem is that I have this regular expression: ^[\\t ]*(?:([.A-Za-z0-9_]+[:]))?(?:[\\t ]*([A-Za-z]{2,4})(?:[\\t ]+(@[A-Za-z0-9_]+(?:(?:\\+|-)[0-9]+)?|\".+?\"|\'.+?\'|[.A-Za-z0-9_]+)(?:[\\t ]*[,][\\t ]*(@[A-Za-z0-9_]+(?:(?:\\+|-)[0-9]+)?|\".+?\"|\'.+?\'|[.A-Za-z0-9_]+))?(?:[\\t ]*[,][\\t ]*(@[A-Za-z0-9_]+(?:(?:\\+|-)[0-9]+)?|\".+?\"|\'.+?\'|[.A-Za-z0-9_]+))?)?)?

for matching some assembler-like string (yeah, I know it's long).

I'm using it like this :

    std::ifstream infile(this->inFile.c_str());
	while (std::getline(infile, line)) {
		lineNumber++;
		line = line.substr(0, line.find("#"));
		if (line == "")
			continue;
		line = reduce(line);

		if (regex_match(line, m, op_reg)) {
			std::vector<std::string> ops;
			ops.push_back(std::to_string(lineNumber));
		    for (int i = 1; i < m.size(); i++) {
		    	if (m[i] != "") {
		    		ops.push_back(m[i]);
		    	}
		    }
		    this->maps.opMap.push_back(ops);
		}
	}

This piece of code is working well on my native Ubuntu installation, but on the WSL version of Ubuntu regex_match always returns false.

Have some of you already encountered this? And if so, how have you managed it ?

[T4lus]

I found the issue is that even in WSL we have \r\n line ending style

[therealkenc]

I found the issue is that even in WSL we have \r\n line ending style

Neither WSL/RealLinux kernel nor the Windows kernel "have \r\n" line ending style. Your input file has a line ending style. This snippet will behave the same on Real Linux and WSL given the same input.

You'll need more code to handle files with differing eol conventions, if that's a design goal. Bonus points for handling all the charset eol conventions, a common one on Windows being UTF16. [Which is needed if your program is going to consume assembler like strings containing characters other than 'merican.]

how have you managed it?

Short answer is strip the \r chars from the ifstream before handing off to std::getline(ins, line, '\n'). Pull the whole file into a std::stringstream is a big hammer way to do that. Boost iostreams are another way to go.

Long answer is use a library like libicu because you can't even assume a [.A-Za-z0-9_] regex has much meaning let alone the file's eol convention. [Shorter long answer: you don't want the long answer.]