Thread: freeing global buffer

I have one of these situations that makes me think I don't really understand something I thought I understood, which it's been true before...I have a global buffer that I want to free() and reuse like this:

Code:

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

char *buffer;

void testfunc (int toggle) {
	char X[]="XXXXXXXXX";
	if (toggle==0) { buffer=malloc(11);
		strcpy(buffer,X);}
	else {buffer=realloc(buffer,strlen(buffer)+11);
		strcat(buffer,X);
	}
}

int main() {
	testfunc(0);
	testfunc(1);
	puts(buffer);
	free(buffer);            // end of 1st iteration
	testfunc(0);
	testfunc(1);
	puts(buffer);
	free(buffer);
	return 0;
}

This code works, but does anyone know what could cause it to produce a double-free/corruption abort at "end of 1st iteration" if:

• THERE IS ONLY ONE free() call previously in the script,
• "buffer" was just malloc'd and realloc'd as above, previous to the free(),
• the total number of bytes allocated to buffer is >0

To me this just does not make sense, it's something I've done many times before (freeing and reusing a malloc'd buffer in a loop) and again, it does not make sense to me, but it keeps happening anyway: I have a global char*, I allocate it memory in a function, it serves it's purpose, then I free it for reuse in main() -- and I get a double free/corruption abort. What factors am I ignorant of?

Have you tried using valgrind to identify the problem.

To me it looks ok, but I haven't got a good test-suite here at home to try it out properly.

--
Mats

Strange.

edit: tested it and runs nicely. Check again (maybe accindentally mixing execution files and running something else?)

Originally Posted by matsp

Have you tried using valgrind to identify the problem.

Mats

No, I'll but perhaps today is the day to try something new, like software...

Originally Posted by slingerland3g

odd if you are really getting that behaviour. Some sort of race condition? Working with this in gcc 4.0.3 on my Ubuntu platform I seem to be ok. Using valgrind would be a good start. Out of curiosity is puts() working as expected as this does have a return variable and happens to be the last call before the free().

Unfortunately I am not having the same problem anymore (after changing the mechanism in the function for another reason) so I may never be able to figure out why (the change was from while recv>0 to while recv>=0, which made some difference to content of buffer, but in either case it was malloc'd and did have some content). w/r/t to puts, yes, it was working which means I should be able to free() the variable!?