[Bug target/17180] [3.4 Regression] nearly all g77 tests fail

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-08-31 06:36 -------
> http://gcc.gnu.org/ml/gcc-patches/2004-07/msg01136.html
> That was easy to find.... (using google).

You're the best :-)  I was desperately searching my mailbox with "davis"...

Anyway, I think the patch is flawed because it does pointer frobbing here:

@@ -388,6 +377,7 @@ malloc_new_inpool_ (mallocPool pool, mal
   void *ptr;
   mallocArea_ a;
   unsigned short i;
+  mallocArea_ *temp;
 
   if (pool == NULL)
     pool = malloc_pool_image ();
@@ -397,11 +387,14 @@ malloc_new_inpool_ (mallocPool pool, mal
 	  || malloc_pool_find_ (pool, malloc_pool_image ()));
 #endif
 
-  ptr = malloc_new_ (s + (i = (MALLOC_DEBUG ? strlen (name) + 1 : 0)));
+  ptr = malloc_new_ (sizeof(mallocArea_*) + s + (i = (MALLOC_DEBUG ? strlen
(name) + 1 : 0)));
 #if MALLOC_DEBUG
   strcpy (((char *) (ptr)) + s, name);
 #endif
   a = malloc_new_ (offsetof (struct _malloc_area_, name) + i);
+  temp = (mallocArea_ *) ptr;
+  *temp = a; 
+  ptr = ptr + sizeof(mallocArea_*);
   switch (type)
     {				/* A little optimization to speed up killing
 				   of non-permanent stuff. */


What happens is that we allocate a new 'ffesymbol' in symbol.c:ffesymbol_new_:

  ffesymbol s;

  s = malloc_new_ks (FFESYMBOL_SPACE_POOL_, "FFESYMBOL", sizeof (*s));


The structure 'ffesymbol' contains the field 'accretes' whose type is defined as
'ffetargetOffset', which is actually 64-bit 'long long' on SPARC.  So the
structure gets 8-byte alignment (BIGGEST_ALIGNMENT is 64). When the system
malloc is invoked, it returns a 8-byte aligned address; but then

ptr = ptr + sizeof(mallocArea_*);

destroys the alignment (ptr==0x439cdc) and the compiler dies because it is
trying to do a 8-byte store to an unaligned address (ptr + 0x48).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=17180