Re: Question on the images in /miniroot

[Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost>]

I wrote:

> dross00%hotmail.com@localhost wrote:
> 
> > The second one doesn't have this problem.  It boots up further without a 
> > delay.
> 
> Okay, then there are some gcc4 specific issue around atari sources.

Ah, no, there is an obvious bug in atari5380.c, which was introduced
between 3.0 and 4.0.

Could you try this one (compiled by gcc4 with the attched patch)?
http://www.ceres.dti.ne.jp/~tsutsui/netbsd/netbsd-atari-BOOT-20080616.gz

---
Index: atari5380.c
===================================================================
RCS file: /cvsroot/src/sys/arch/atari/dev/atari5380.c,v
retrieving revision 1.41
diff -u -r1.41 atari5380.c
--- atari5380.c 10 May 2006 06:24:02 -0000      1.41
+++ atari5380.c 16 Jun 2008 15:34:08 -0000
@@ -138,14 +138,14 @@
 
 #define        set_scsi_dma(addr, val) (void)(                                 
\
        {                                                               \
-       u_char  *address = (u_char*)__UNVOLATILE(addr+1);               \
+       volatile u_char *address = (volatile u_char *)addr + 1;         \
        u_long  nval     = (u_long)val;                                 \
        __asm("movepl   %0, %1@(0)": :"d" (nval), "a" (address));       \
        })
 
 #define        get_scsi_dma(addr, res) (                                       
\
        {                                                               \
-       u_char  *address = (u_char*)__UNVOLATILE(addr+1);               \
+       volatile u_char *address = (volatile u_char *)addr + 1;         \
        u_long  nval;                                                   \
        __asm("movepl   %1@(0), %0": "=d" (nval) : "a" (address));      \
        res = (u_long)nval;                                             \
---
Maybe it's much better to rewrite them to use inline functions, though.

---
Izumi Tsutsui