[splint-discuss] Splint and malloc

[Phil Longstaff]

On February 23, 2009 04:35:48 am Roland Illig wrote:
> Phil Longstaff schrieb:
> > Is there any way of indicating to splint that if a struct with pointers
> > is malloc'ed and the pointers are given values, that the previous values
> > don't need to be freed because it is new memory?
> >
> > Example:
> >
> > typedef struct {
> >
> > char* x;
> >
> > } struct_t;
> >
> > struct_t* s = malloc( sizeof(struct_t) );
> >
> > s.x = malloc(10);
> >
> > will give a warning that s.x is implied only and hasn't been freed,
> > leading to memory leak. However, since s is pointing to new storage,
> > there isn't anything to be freed.
>
> The warning wants to tell you that you need to free s->x (not s.x)
> later. This works for me:
>
> #include <assert.h>
> #include <stdlib.h>
>
> typedef struct {
>         char* x;
> } struct_t;
>
> int main(void)
> {
>         struct_t* s;
>
>         s = malloc( sizeof(struct_t) );
>         assert(s != NULL);
>         s->x = malloc(10);
>
>         /* ... */
>
>         free(s->x);
>         free(s);
>
>         return 0;
> }

Yes, I meant s->x not s.x.  I find that I'm asking the wrong question.

	s = malloc(sizeof(struct_t))
	s->x = ...

produces no warnings, whereas

	s = g_malloc(sizeof(struct_t))
	s->x = ...

produces the warning that implicitly only storage s->x is not freed before the 
assignment.  g_malloc() is the version of malloc included in glib on linux 
(library.gnome.org/devel/glib/). Maybe splint knows about malloc and that it 
always returns fresh storage.  Is there any way to tell it that g_malloc also 
returns fresh storage (the same way I can use -booltype to tell it that 
gboolean is a boolean type)?

Phil
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cs.virginia.edu/pipermail/splint-discuss/attachments/20090223/e5fdf859/attachment-0002.html