Thread: regfree() failure

I wanted to write a function that would accept a string and a regular expression, then fill in a global struct with the beginning and end positions of a match, and return the actual match as another character string. Because regcomp() from regex.h only finds the first match in a line, I then wanted to use a static char copy of the original string which could be "blanked out" and matched against again to find multiple matches per line.

As far as I can tell, something about including the static char causes regfree() to fail, such that regcomp acts like it's regmatch_t structure has not been freed from the last call (so matches nothing). I say this because an alternate version of the following script works fine (in which there is no static char in the "regexp_match" function, and the string to match must be "blanked out" before it is fed back into the function again. But I would prefer it to work this way and can't see what the problem is!

Code:

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

char *chomp (char *line); // removes *nix newline from string
char *regexp_match (char *string, char *regexp);
 
struct matchspec {
	int begin;
	int end;
} MSpec;


int main () {
	int i, count=1;
	char input[164], *regexp, *string, *match;
	
	puts("character string:");
	fgets(input,164,stdin);
	string=chomp(input);

	while (1) {
		regexp = NULL;
		puts("regular expression:");
		fgets(input,164,stdin);
		if (input[0] == '\n') return;
		regexp=chomp(input);
		printf("input: \"%s\" regexp: \"%s\"\n", string, regexp); 
		if ((match=regexp_match(string,regexp)) != NULL) {
	     		printf("%d -- B:%d E:%d is \"%s\"\n", count,MSpec.begin,MSpec.end,match); 
				// retain value of static char in regexp_match by using NULL	
			while ((match=regexp_match(NULL,regexp)) != NULL) {
				count++;
				printf("%d -- B:%d E:%d is \"%s\"\n", count,MSpec.begin,MSpec.end,match); 	
			}
		} else puts("No match.");
		free(regexp);
	}
}


char *chomp (char *line) {	// a standard function
	int len, elen;
	char *chompd, *end;
	len = strlen(line);
	end = strrchr(line,10);
	if (end == NULL) return line;
	else { 	elen = strlen(end);
		if (elen > 1) return line;
		else {	chompd = (char *)malloc(len-1);
			strncpy(chompd,line,(len-1));
			return chompd;
		}
	}
}


char *regexp_match (char *string, char *regexp) {
	int i, w=0, len;
	char *word = NULL;
	static char *copy;
	regex_t rgT;
	regmatch_t match;
	regcomp(&rgT,regexp,REG_EXTENDED);
	if (string != NULL) {		// write "string" to the static char ("copy")
		len=strlen(string);	// not needed in alternate version
		copy=(char *)malloc(len+1);
		strcpy(copy,string);
	} 
	if ((regexec(&rgT,copy,1,&match,0)) == 0) {
		MSpec.begin = (int)match.rm_so;
		MSpec.end = (int)match.rm_eo;
		len = MSpec.end-MSpec.begin;
		word=(char *)malloc(len+1);
		for (i=MSpec.begin; i<MSpec.end; i++) {
			word[w] = copy[i];
			w++; }
		word[w]=0;
			// blank out match in "copy" for next call
		for (i=MSpec.begin; i<MSpec.end; i++) {
			copy[i]=32;}
	}
	regfree(&rgT); 	// DOESN'T WORK !!!
	return word;
}

The output goes like this:

Code:

character string:
the 6th one
regular expression:
th
input: "the 6th one" regexp: "th"
1 -- B:0 E:2 is "th"
2 -- B:5 E:7 is "th"
regular expression:
th
input: "the 6th one" regexp: "th"
No match.

Which actually means that regfree() worked the first time...it's not until "copy" is reinitialized with a new string that the problem begins.

Does the compiler not complain about this in main?

Code:

if (input[0] == '\n') return;

You should return an integer in main...

Sorry, I can't see anything directly wrong in your usage of regexp handling.

--
Mats