Replace regular expressions for performance

[probonopd]

Based on the results of AppImageCommunity/appimaged#53, regular expressions seem to be really CPU intensive.

Hence, we should replace them with something more efficient. Possibly fnmatch(3) would be more efficient (to be verified),

After having replaced regular expressions with something else, please test that

• --appimage-extract still works with * wildcards
• appimaged consumes significantly less CPU

[TheAssassin]

I've run the following two test programs in a VM (4GiB RAM, 4 CPU cores).

#include <stdio.h>
#include <fnmatch.h>
#include <stdlib.h>

int main(int argc, char* argv[]) {
    if (argc != 2) {
        printf("Usage: %s <num rounds>\n", argv[0]);
	return 2;
    }

    const int rounds = atoi(argv[1]);

    char pattern[] = "abc*nop";
    char test[] = "abcdefghijklnmop";

    for (int i = 0; i < rounds; ++i) {
        fnmatch(pattern, test, FNM_PATHNAME);
    }

    return 0;
}

#include <stdio.h>
#include <regex.h>
#include <stdlib.h>

int main(int argc, char* argv[]) {
    if (argc != 2) {
        printf("Usage: %s <num rounds>\n", argv[0]);
	return 2;
    }

    const int rounds = atoi(argv[1]);

    char pattern[] = "abc.*nop";
    char test[] = "abcdefghijklnmop";

    for (int i = 0; i < rounds; ++i) {
        regex_t regex;
        regcomp(&regex, pattern, /*REG_ICASE |*/ REG_EXTENDED);
        regmatch_t matches[1];
        regexec(&regex, test, 1, matches, 0);
	regfree(&regex);
    }

    return 0;
}

The fnmatch based implementation is a lot shorter than the regex based one.

In terms of performance, these are the results for 10.000.000 iterations (both applications compiled using gcc -O0 -o <app> <app>.c):

$ time ./regex 10000000
real 1m54.342s
...

$ time ./fnmatch 10000000
real 0.780s

With -O3, the results change to 1m44.257s and 0.738s.