In article <78a2305a05073006481c030de8@mail.gmail.com>
acruhl@gmail.com wrote:

> Has anyone built from the netbsd-3 branch?
> 
> I just tried this morning to cross build a dreamcast release on my
> amd64 running 3.0_BETA, and it failed with this:
 :
> insn-emit.c -o insn-emit.o.tmp
> insn-emit.c: In function `gen_mshfhi_l_di':
> insn-emit.c:5039: warning: integer constant is too large for "long" type
> insn-emit.c: In function `gen_mshflo_l_di':
> insn-emit.c:5057: warning: this decimal constant is unsigned only in ISO C90

autobuild on releng seems to have no problem:
http://releng.netbsd.org/builds/netbsd-3/200507250000Z/
but insn-emit.c is a generated file on the host, so maybe
few people tried built on 64bit hosts and sh.md might have
LP64 problem. Could you show the lines which cause the warnings?
How about this diff taken from gcc cvsweb?
---

Index: gnu/dist/gcc/gcc/config/sh/sh.md
===================================================================
RCS file: /cvsroot/src/gnu/dist/gcc/gcc/config/sh/sh.md,v
retrieving revision 1.2
diff -u -r1.2 sh.md
--- gnu/dist/gcc/gcc/config/sh/sh.md	29 May 2004 00:49:15 -0000	1.2
+++ gnu/dist/gcc/gcc/config/sh/sh.md	31 Jul 2005 11:39:54 -0000
@@ -1981,7 +1981,7 @@
   [(const_int 0)]
   "
 {
-  if (INTVAL (operands[2]) == (unsigned) 0xffffffff)
+  if ((unsigned)INTVAL (operands[2]) == (unsigned) 0xffffffff)
     emit_insn (gen_mshflo_l_di (operands[0], operands[1], CONST0_RTX (DImode)));
   else
     emit_insn (gen_mshfhi_l_di (operands[0], CONST0_RTX (DImode), operands[1]));

---
Izumi Tsutsui