C++ std::regex segmentation fault - regex

Question

I have various data that I need to parse and get the weight out of it.
I'm using
C++11
std::regex
Debian 9.9
gcc 6.3.0
The problem is that sometimes segmentation fault occurs, it happens very rarely.
The input that throws the error mostly consist of just space and newline characters.
Here is the regex:
(?:\b(?:(kilogram\.*s*\.*|kg\.*s*\.*)(?:[^[:alnum:]])*)(?:\s*weight\s*)*(?:\s*is\s*|\s*are\s*)*)\W*([\d\.,]*\d+\b)|(?:(?:[\s\.]?|^)([\d\.,]*\d+)\W*(kilogram\.*s*\.*|kg\.*s*\.*)\b)
Example regex that works on regex101.com but throws segmentation fault in C++ on my Debian server regex101
Here are some more regex101 examples of input, just to fast get an idea of what regex is searching for.
Here is an example of C++ code that fails.
And here is the same C++ code that works, but using another online compiler (cpp.sh).
Can someone please help me to solve this segmentation fault problem?
Thank you.

Answer 1

I have the same issue with a simple regex .+ and [a-zA-Z0-9\\+/=]+.
I have tried different compilers: g++, clang++, clang-cl on Windows, and g++, clang++ on Linux (WSL).
On Windows, the application freeze and ends suddenly. On Ubuntu (WSL), I have the Segmentation Fault.
The error happens for g++ on Windows with c++11, c++14, c++17 and also c++20.
Limit
In your example, your data regex101 has 31275 characters which, I suppose, is too many for regex_match.
Here is the program I used to guess the maximal length of the data.
#include <iostream>
#include <regex>
int main(int argc, char **argv) {
int length = argc > 1 ? std::stoi(std::string(argv[1])) : 30000;
std::regex testRegex(".+");
std::string data = "";
for (int i = 0; i < length; ++i) {
data += "a";
}
std::cout << "Match: " << std::regex_match(data, testRegex) << std::endl;
return 0;
}
// Limit before crash (it's a bit random so the limit is not accurate)
// Windows 11
// clang++ Windows : 4999998
// clang-cl Windows : 4999998
// g++ Windows : 6833
// WSL Ubuntu 20.04
// clang++ WSL : 23804
// g++ WSL : 26187
How to solve
According to this test, data has a size limit, and the application will stop if the limit is exceeded.
What you can do is:
Remove some unnecessary spaces before using regex_match
Split the data in half
On Windows, you can use clang++ to increase the limit to 5M chars
For me, I split my data in half because the regex [a-zA-Z0-9\\+/=]+ doesn't require the entire input.
If anybody knows how we can increase the limit (with some flags or #define), I am interested.