https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125730
--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Already tree_to_aff_combination ((char * * *) (((unsigned long) first_kw_arg_12
+ (unsigned long) values_16(D)) - (unsigned long) argnames_11(D)), char ***)
strips the integer conversions which is because it uses STRIP_NOPS which
happily does. Of course this irrevocably drops important information we cannot
conservatively recover (similar for signed/unsigned details) :/
The following fixes the testcase, selecting
_20 = (sizetype) first_kw_arg_12;
_26 = (sizetype) argnames_11(D);
_27 = _20 - _26;
# PT = nonlocal null
_28 = values_16(D) + _27;
MEM[(char * * *)_28 + ivtmp.11_8 * 1] = fetched_value.1_6;
but I fear some "massive" fallout.
diff --git a/gcc/tree-affine.cc b/gcc/tree-affine.cc
index b5b7249c675..a55074ba7b4 100644
--- a/gcc/tree-affine.cc
+++ b/gcc/tree-affine.cc
@@ -32,6 +32,7 @@ along with GCC; see the file COPYING3. If not see
#include "dumpfile.h"
#include "cfgexpand.h"
#include "value-query.h"
+#include "tree-ssa.h"
/* Extends CST as appropriate for the affine combinations COMB. */
@@ -312,8 +313,7 @@ expr_to_aff_combination (aff_tree *comb, tree_code code,
tree type,
tree itype = TREE_TYPE (inner);
enum tree_code icode = TREE_CODE (inner);
- /* STRIP_NOPS */
- if (tree_nop_conversion_p (otype, itype))
+ if (useless_type_conversion_p (otype, itype))
{
tree_to_aff_combination (op0, type, comb);
return true;
@@ -394,7 +394,7 @@ tree_to_aff_combination (tree expr, tree type, aff_tree
*comb)
machine_mode mode;
int unsignedp, reversep, volatilep;
- STRIP_NOPS (expr);
+ STRIP_USELESS_TYPE_CONVERSION (expr);
code = TREE_CODE (expr);
switch (code)
