gEDA: gEDA/gaf 20050313 almost ready, but not released...

[Ales Hvezda <ahvezda@xxxxxxxxxxxxxx>]

Hi All,

	I was doing my last bits of cleanup / additions before doing 
the release (mainly symbols) when I observed some rather unpleasant
behavior.  Basically gschem upon exit would hang.  Looking in
the debugger, gschem was getting stuck in a malloc call caused by
a g_strdup_printf (in o_undo.nw:o_undo_cleanup).  Usually when this
happens it points to the heap that is seriously messed up.  I don't think
that the undo code is the problem.

	So, I ran valgrind against gschem and it generated the following
error (in a different part of the code):

==31275== Invalid free() / delete / delete[]
==31275==    at 0x1B907460: free (vg_replace_malloc.c:153)
==31275==    by 0x1BBBB9F3: g_free (in /usr/lib/libglib-2.0.so.0.600.1)
==31275==    by 0x1BBA9289: (within /usr/lib/libglib-2.0.so.0.600.1)
==31275==    by 0x1BBA8E8A: (within /usr/lib/libglib-2.0.so.0.600.1)
==31275==  Address 0x1C69F980 is 0 bytes inside a block of size 228 free'd
==31275==    at 0x1B907460: free (vg_replace_malloc.c:153)
==31275==    by 0x1BBBB9F3: g_free (in /usr/lib/libglib-2.0.so.0.600.1)
==31275==    by 0x1B92BD9A: delete_font_set (o_text_basic.nw:1491)
==31275==    by 0x1BBA8DB5: (within /usr/lib/libglib-2.0.so.0.600.1)


	Usually when valgrind complains, it is a legit problem.  
The code in question is:

static gboolean
delete_font_set (gpointer key, gpointer value, gpointer user_data)
{
  OBJECT *tmp = (OBJECT*)value;
  TOPLEVEL *toplevel = (TOPLEVEL*)user_data;

  if (tmp != NULL) {
    if (tmp->font_prim_objs != NULL) {
      s_delete_list_fromstart (toplevel, tmp->font_prim_objs);
      tmp->font_prim_objs = NULL;
    }
    /* do not use s_delete() as tmp is not fully initialized */
    g_free (tmp->name);
    g_free (tmp);   <---- This is what valgrind is complaining about.
   
  }

  return TRUE;
}

Do we really need to free the hash value itself inside of the delete
function?

I'll keep looking at this (and run efence), but in the mean time,
I don't want to release with such a glaring problem. Sorry. ;-( 
I'm hoping that this is the cause of the malloc hang.

								-Ales