C++ (C++11) regular expressions differences on OS X and Linux

Question

I'm trying to get my code to work on both OS X and Linux the same. The code below is compiled with clang++ --std=c++11 regextest.cpp

#include <regex>
#include <iostream>
int main()
{
    std::string str = "/api/asd/";
    std::string pattern = "/api/(.*)/";
    std::cout << "Starting matching" << std::endl;
    std::smatch matches;
    if (std::regex_match(str, matches, std::regex(pattern, std::regex::egrep)))
    {
        std::cout << "Found match!" << std::endl;
        std::cout << "All matches: ";
        for (auto& it : matches)
            std::cout << it << ", ";
        std::cout << std::endl;
    }
    return 0;
}

On OS X, the result of running this code is:

Starting matching
Found match!
All matches: /api/asd/, asd,

On Linux, on the other hand (Gentoo, libstdc++ 3.3)

Starting matching
Found match!
All matches: /api/asd/, /asd/, //

How does it match /api/ on Linux? Why?

Additionally, trying to use a pattern like /api/([^/]) fails completely in Linux and matches nothing but works well in OS X.

I've tried many combinations of match types, (basic, extended, grep, egrep, awk) with escaped and unescaped ( and ) (depending on the match type) and nothing produces the expected results on Linux.

Looks like theres something wrong with your compiler on Linux. Try using GCC, and make sure its updated. — Havenard, Jul 21 '15 at 21:09
The libstdc++ version that shipped with g++-4.9 was the first to support ``, see https://stackoverflow.com/a/12665408/241631. My guess is you're using an older version. — Praetorian, Jul 21 '15 at 21:12
Have you tried using `std::regex::ECMAScript` instead of `std::regex::egrep`? I'm not sure what are the peculiarieties of egrep's regular expressions, but ECMAScript's should work fine. — Havenard, Jul 21 '15 at 21:12
I found regex in the std (in linux) to behave differently than boost regex, so I think std regex has some bugs, in the past at least. — alfC, Jul 22 '15 at 02:27
@n.m. gcc --version: `gcc (Gentoo 4.8.4 p1.4, pie-0.6.1) 4.8.4` — Irenej Marc, Jul 22 '15 at 13:02
I don't understand, though, why matching works at all, but only seemingly incorrectly if there is absolutely no regex support. — Irenej Marc, Jul 22 '15 at 13:03

score 1 · Answer 1 · answered Jul 22 '15 at 13:54

1

As suggested by the comments, this issue was solved by upgrading gcc to 4.9. (~amd64 flag currently required to do this on Gentoo).

answered Jul 22 '15 at 13:54

Irenej Marc

287
3
9

C++ (C++11) regular expressions differences on OS X and Linux

1 Answers1