[RFC]: patch for PR c/22476 -Wmissing-format-attribute

[Kaveh R. Ghazi]

 > [...] However it fails to find the initialization case as expected
 > in testcase miss-3.c in the patch below.
 >
 > For that testcase, GCC never calls 'convert_for_assignment'.
 > Instead it leaves 'digest_init' through another clause and returns
 > 'inside_init'.  What I'm wondering is whether 'digest_init' needs
 > an additional call to 'convert_for_assignment' on 'inside_init' in
 > the relevant clause or if I have to duplicate the missing attribute
 > detection code in that clause in 'digest_init'?

In case that wasn't clear, I mean this additional hunk.  If I add this
in, the miss-3.c testcase yields the expected diagnostics.  But I
don't know if this is correct in general.

		Thanks,
		--Kaveh

diff -rup orig/egcc-CVS20050714/gcc/c-typeck.c egcc-CVS20050714/gcc/c-typeck.c
--- orig/egcc-CVS20050714/gcc/c-typeck.c	2005-07-05 20:01:52.000000000 -0400
+++ egcc-CVS20050714/gcc/c-typeck.c	2005-07-15 09:13:47.854507882 -0400
@@ -4442,7 +4472,8 @@ digest_init (tree type, tree init, bool 
 	  inside_init = error_mark_node;
 	}
 
-      return inside_init;
+      return convert_for_assignment (type, inside_init, ic_init, NULL_TREE,
+				     NULL_TREE, 0);
     }
 
   /* Handle scalar types, including conversions.  */