Subject: Re: Need advice how to debug VM problem on port-hp700.
To: None <port-hp700@NetBSD.org>
From: Jochen Kunz <jkunz@unixag-kl.fh-kl.de>
List: port-hp700
Date: 10/29/2003 23:11:27
On 2003.10.29 21:26 Charles M. Hannum wrote:

> The way this works, it would have to be a gas bug, not a gcc bug.
Well, for me a compiler is a black box where I put a .c file into one
end and a .o or ready-to-execute file comes out of the other end. ;-)

Comparing the assembler output of the compiler for non broken:
        .LEVEL 1.1
        .section .ident ; .asciz "$NetBSD: nfs_boot.c,v 1.59 2003/06/29
22:32:14 fvdl Exp $" ; .previous
        .globl my_debug_test_2
        .data
        .align 4
        .type   my_debug_test_2, @object
        .size   my_debug_test_2, 4
my_debug_test_2:
        .word   0
        .globl nfs_boot_rfc951
        .align 4
        .type   nfs_boot_rfc951, @object
        .size   nfs_boot_rfc951, 4
nfs_boot_rfc951:
        .word   1
        .globl nfs_boot_bootparam
        .align 4
        .type   nfs_boot_bootparam, @object
        .size   nfs_boot_bootparam, 4
nfs_boot_bootparam:
        .word   1
        .section        .rodata


and broken file:
        .LEVEL 1.1
        .globl my_debug_nfs_boot
        .data
        .align 4
        .type   my_debug_nfs_boot, @object
        .size   my_debug_nfs_boot, 4
my_debug_nfs_boot:
        .word   0
        .section .ident ; .asciz "$NetBSD: nfs_boot.c,v 1.59 2003/06/29
22:32:14 fvdl Exp $" ; .previous
        .globl my_debug_test_2
        .align 4
        .type   my_debug_test_2, @object
        .size   my_debug_test_2, 4
my_debug_test_2:
        .word   0
        .globl nfs_boot_rfc951
        .align 4
        .type   nfs_boot_rfc951, @object
        .size   nfs_boot_rfc951, 4
nfs_boot_rfc951:
        .word   1
        .globl nfs_boot_bootparam
        .align 4
        .type   nfs_boot_bootparam, @object
        .size   nfs_boot_bootparam, 4
nfs_boot_bootparam:
        .word   1
        .section        .rodata

With a quick glance into the gas manual, yes, this is a gas bug.
".previous" does not work correct. The effect on find_port from
sys/netinet/ip_h323_pxy.c is the same. 
-- 


tschüß,
       Jochen

Homepage: http://www.unixag-kl.fh-kl.de/~jkunz/