[Bug c/37985] [4.4/4.5/4.6/4.7 Regression] unsigned char shift lacks "statement with no effect" warning

["rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37985

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |

--- Comment #10 from Richard Guenther <rguenth at gcc dot gnu.org> 2012-01-16 09:00:15 UTC ---
The patch at http://gcc.gnu.org/ml/gcc-patches/2012-01/msg00669.html does
not work as we then warn for gcc.dg/20040202-1.c

gcc.dg/20040202-1.c:7:5: warning: statement with no effect [-Wunused-value]

which we do because we fold the memcpy (dest, src, 0); to
(dest = src, (void *)dest) with a TREE_NO_WARNING flag on the cast.

So we run afoul of having only one TREE_NO_WARNING flag, thus trying to
disable a specific warning using a cast will also disable other useful
ones.

The patch that caused this regression
(http://gcc.gnu.org/ml/gcc-patches/2006-08/msg01041.html) is patching a too
generic place to handle the situation.

That issue is probably similar to the fallout of the above patch - we fold
too much and too early for these kind of warnings (thus the C AST deviates
too much from the source input).

For the statement with no effect warning we emit that from c_process_expr_stmt
()
which calls c_fully_fold _before_ emitting those warnings (for whatever
historical reasons...).  Doing

Index: gcc/c-typeck.c
===================================================================
--- gcc/c-typeck.c      (revision 183205)
+++ gcc/c-typeck.c      (working copy)
@@ -9186,8 +9190,6 @@ c_process_expr_stmt (location_t loc, tre
   if (!expr)
     return NULL_TREE;

-  expr = c_fully_fold (expr, false, NULL);
-
   if (warn_sequence_point)
     verify_sequence_points (expr);

@@ -9213,6 +9215,8 @@ c_process_expr_stmt (location_t loc, tre
       || TREE_CODE (exprv) == ADDR_EXPR)
     mark_exp_read (exprv);

+  expr = c_fully_fold (expr, false, NULL);
+
   /* If the expression is not of a type to which we cannot assign a line
      number, wrap the thing in a no-op NOP_EXPR.  */
   if (DECL_P (expr) || CONSTANT_CLASS_P (expr))

ontop of the patch "fixes" this issue.