[patch,fortran,committed] Fixed memory allocation in expr.c for 4.2 [PR29713] (was: Re: make check: Regressions on trunk and 4.2)

[Tobias Burnus <burnus at net-b dot de>]

Hi,

Tobias Burnus wrote:
>> Line 744 of gcc-4_2-branch/gcc/fortran/decl.c is function
>> gfc_set_constant_character_len:
>>
>>       gfc_free (expr->value.character.string);
>>
>> This looks good in principle and also matches 4.3. Any ideas what goes
>> wrong here?
>>     
I should trust valgrind more and do no step-wise debugging in gdb with
"-O2".

The problem is PR29713, which I found on the trunk and which FX nicely
debugged and fixed
(but unfortunately only on the trunk).

I somehow believed that this was fixed in GCC 4.2 in the meanwhile
(especially since gfc_set_constant_character_len looked the same). But
of cause the problem was in gfc_simplify_expr as valgrind told me.

I followed FX by submitting it as obvious. (Since the testcase is
already in gcc, I didn't included a reduced testcase.)

Same fix needs to be done for 4.1 as
http://gcc.gnu.org/viewcvs/branches/gcc-4_1-branch/gcc/fortran/expr.c?view=markup
shows.

Tobias

--- branches/gcc-4_2-branch/gcc/fortran/expr.c	2006/11/09 10:26:10	118619
+++ branches/gcc-4_2-branch/gcc/fortran/expr.c	2006/11/09 14:28:59	118620
@@ -1485,7 +1485,7 @@
 	  gfc_extract_int (p->ref->u.ss.start, &start);
 	  start--;  /* Convert from one-based to zero-based.  */
 	  gfc_extract_int (p->ref->u.ss.end, &end);
-	  s = gfc_getmem (end - start + 1);
+	  s = gfc_getmem (end - start + 2);
 	  memcpy (s, p->value.character.string + start, end - start);
 	  s[end-start+1] = '\0';  /* TODO: C-style string for debugging.  */
 	  gfc_free (p->value.character.string);