[PATCH] fix for ldb_msg_add_string() to increase value lifetime

[Tim Potter]

On Mon, 2007-09-03 at 21:48 +1000, Andrew Bartlett wrote:
> On Mon, 2007-09-03 at 21:35 +1000, Tim Potter wrote:
> > Hey ldb people - I would like to propose this patch to take a copy of
> > the string value when calling ldb_msg_add_string().  It's actually
> > pretty easy to find garbage in your attribute values if the lifetime of
> > the const char *str parameter is less than the lifetime of the
> > ldb_message.
> > 
> > Comments?  It looks OK to me.  (-:
> 
> The main problem is that much of the current code assumes the current
> lifetime semantics, and we don't want too much copies of the string
> around 'just in case'.  We should also watch for failures of the memory
> allocation. 

Good point about the memory allocation failure.  I'll update the patch.

However it's hardly keeping an extra copy "just in case".  It's keeping
an extra copy to avoid data corruption or crashes!  I think some extra
memory is worth the price of ensuring correctness.


Tim.