[haiku-commits] Re: haiku: hrev43836 - src/build/libroot headers/build/host/darwin/sys
- From: John Scipione <jscipione@xxxxxxxxx>
- To: haiku-commits@xxxxxxxxxxxxx
- Date: Fri, 9 Mar 2012 15:30:47 -0500
On Fri, Mar 9, 2012 at 3:03 PM, Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> wrote:
> Well, I left in the problematic code in my reply on purpose, but here it
> is again:
>
> > + dirpath = (char *)malloc(MAXPATHLEN);
> > + if (dirpath == NULL) {
> > + // ran out of memory allocating dirpath
> > + errno = ENOMEM;
> > + return -1;
> > + }
> > +
> > + if (fcntl(fd, F_GETPATH, dirpath)< 0) {
> > + // failed to get the path of fd, fcntl sets errno
>
> Leak of dirpath here.
>
>
> > + return -1;
> > + }
> > +
> > + if (strlcat(dirpath, path, MAXPATHLEN)> MAXPATHLEN) {
> > + // full path is too long, set errno and return
> > + errno = ENAMETOOLONG;
>
> Leak of dirpath here.
>
I am not trying to be difficult but I still don't see where the memory leak
is. You are looking at old code from an earlier commit. The final code is
as follows:
// fullPath is allocated
char* fullPath = (char *)malloc(MAXPATHLEN);
if (fullPath == NULL) {
// ran out of memory allocating dirpath
errno = ENOMEM;
return -1;
}
// if malloc fails, no memory leak. If malloc succeeds fullPath is passed
into get_path()
if (get_path(fd, path, &fullPath) < 0) {
free(fullPath);
return -1;
}
// if get_path fails, memory is freed, if not the next code is run
int status = mkdir(fullPath, mode);
free(fullPath);
return status;
// fullPath is used, memory is freed, function returns.
fullPath is allocated before entering get_path and free after so get_path
is not responsible for freeing fullPath. It either returns -1 for failure
or 0 for success and fullPath is freed afterwards.
John Scipione
Other related posts:
