FreeTDS Development Group

["James K. Lowden" <jklowden AT schemamania.org>]

On 31 Jan 2004 16:22:16 +0100, Frediano Ziglio <freddyz77 AT tin.it> wrote:
> Il gio, 2004-01-29 alle 22:28, Edward Quackenbush ha scritto:
> > I encountered a memory leak when repeatedly opening and closing a 
> > conection using the DB-lib interface.  It appears that the
> > dbstring_free function in dblib.c does not free the string.  The
> > following patch should correct this issue. 
> > eq 
> > 
> 
> Applied

I don't see why this is needed.  I think the pre-patched code is right. 
Can someone explain why not?  

struct dbstring
{
        BYTE *strtext;
        DBINT strtotlen;
        struct dbstring *strnext;
};
typedef struct dbstring DBSTRING;

 1  dbstring_free(DBSTRING ** dbstrp)
 2  {
 3          if ((dbstrp != NULL) && (*dbstrp != NULL)) {
 4                  if ((*dbstrp)->strnext != NULL) {
 5                          dbstring_free(&((*dbstrp)->strnext));
 6                  }
 7                  free(*dbstrp);
 8                  *dbstrp = NULL;
 9          }
10  }

So, if I have:

DBSTRING a, b;

a.strtext = malloc(10);
a.strtotlen = 10;
a.strnext = &b;

b.strtext = malloc(10);
b.strtotlen = 10;
b.strnext = NULL;

Then the call dbstring_free(&a) should:

== nest level 0, examining a ==
#3  dbstrp != NULL) && (*dbstrp != NULL), so
#4  (*dbstrp)->strnext != NULL // because it's b, so
#5  dbstring_free(&((*dbstrp)->strnext));
== nest level 1, examining b ==
#3  dbstrp != NULL) && (*dbstrp != NULL), so
#4  (*dbstrp)->strnext == NULL, so
#7  free(*dbstrp); // frees b
#8  *dbstrp = NULL;
#10 return
== nest level 0, continuing with a ==
#7  free(*dbstrp); // frees a
#8  *dbstrp = NULL;
#10 return

Thus, both a and b are freed.  The effect of the patch should be nil,
calling free(3) on a NULL pointer, set by the recursive call.  

--jkl