Do I need to call free() after a failed getline()?

Question

The following block causes a memory leak:

FILE *fp = fopen(path, "r");
char *line = NULL;
size_t len = 0;
ssize_t read = -1;

while ((read = getline(&line, &len, fp)) != -1) {
        /*Do something*/
}

120 bytes in 1 blocks are definitely lost...

...getline (getline.c:34)

I can fix this by adding a free():

while ((read = getline(&line, &len, fp)) != -1) {
        /*Do something*/
}
free(line);

My question is: Why does getline allocate memory for line when it fails? And why do I not need to free(line) for every call to getline?

Answer 1

The setup is such that you can pass a previously allocated block of memory to getline(), and it will allocate more (realloc()) if it needs it. (Or you can start with no memory allocated, as here.) It can report failure or EOF, but doesn't release the space that was allocated — hence you need to free it. If the file is an empty file and you start off with no data, you might not get any space allocated; on the other hand, it may have allocated some space before it attempts to get data from the file.

But you know that if the line pointer is not null, it was allocated and must be freed.

Answer 2

getline can fail in several different ways, and may or may not have realloced the memory one or more times when it fails. To be consistent, it never frees the memory -- see one possible implementation at http://www.opensource.apple.com/source/cvs/cvs-19/cvs/lib/getline.c

Such consistency makes it easier both for the implementation and for the caller ... you just always free the line rather than having to check the error code to see whether to free it.