[gfortran,committed] Fix PR31974: mismatched types in multiplication

[FX Coudert]

Hi all,

I committed the patch below after bootstrapping and regtesting on  
x86_64-linux. It fixes PR31974, a 4.3 regression due to a bug exposed  
by my recent malloc/free patch. The front-end generated mismatched  
types in the calculation of the size of the memory to allocate.

FX




2007-05-19  Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>

	PR fortran/31974
	* trans-array.c (gfc_trans_auto_array_allocation): Avoid
	multiplication of mismatched types.

2007-05-19  Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>

	PR fortran/31974
	* gfortran.dg/char_allocation_1.f90: New test.



Index: fortran/trans-array.c
===================================================================
--- fortran/trans-array.c       (revision 124856)
+++ fortran/trans-array.c       (working copy)
@@ -3846,7 +3846,8 @@ gfc_trans_auto_array_allocation (tree de
    /* The size is the number of elements in the array, so multiply  
by the
       size of an element to get the total size.  */
    tmp = TYPE_SIZE_UNIT (gfc_get_element_type (type));
-  size = fold_build2 (MULT_EXPR, gfc_array_index_type, size, tmp);
+  size = fold_build2 (MULT_EXPR, gfc_array_index_type, size,
+                     fold_convert (gfc_array_index_type, tmp));
    /* Allocate memory to hold the data.  */
    tmp = gfc_call_malloc (&block, TREE_TYPE (decl), size);
Index: testsuite/gfortran.dg/char_allocation_1.f90
===================================================================
--- testsuite/gfortran.dg/char_allocation_1.f90 (revision 0)
+++ testsuite/gfortran.dg/char_allocation_1.f90 (revision 0)
@@ -0,0 +1,11 @@
+! PR fortran/31974
+! { dg-do run }
+  subroutine foo (n)
+    integer :: n
+    character (len = n) :: v(n)
+    v = ''
+    if (any (v /= '')) call abort
+  end subroutine foo
+
+  call foo(7)
+  end