GCC build failed for native with your patch on 2003-04-17T21:46:51Z.

[Kazu Hirata]

Hi Geoff,

> Among other things, I don't see why a NE expression will necessarily
> have a SET_DEST:

I tried powerpc-eabisim from Athlon, but I wasn't able to reprocuce
youe problem.  So here is a wild guess.  Does the patch fix the
problem?

Thanks,

Kazu Hirata

Index: reload1.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/reload1.c,v
retrieving revision 1.389
diff -u -r1.389 reload1.c
--- reload1.c	17 Apr 2003 21:42:19 -0000	1.389
+++ reload1.c	18 Apr 2003 04:28:15 -0000
@@ -9307,10 +9307,11 @@
 	      && GET_CODE (cnd) == NE
 	      && GET_CODE (XEXP (cnd, 0)) == REG
 	      && SCALAR_INT_MODE_P (GET_MODE (XEXP (cnd, 0)))
+	      && HARD_REGNO_NREGS (REGNO (XEXP (cnd, 0)), GET_MODE (XEXP (cnd, 0))) == 1
 	      && GET_CODE (XEXP (cnd, 1)) == CONST_INT)
 	    {
 	      rtx implicit_set =
-		gen_rtx_SET (VOIDmode, SET_DEST (cnd), SET_SRC (cnd));
+		gen_rtx_SET (VOIDmode, XEXP (cnd, 0), XEXP (cnd, 1));
 	      move2add_note_store (SET_DEST (implicit_set), implicit_set, 0);
 	    }
 	}