Subject: current kernel GENERIC build failure, with optimization specified
To: NetBSD Current Users <>
From: Barry Bouwsma <>
List: current-users
Date: 12/25/2004 11:41:53

Latest NetBSD-current downloaded a couple days ago.  I've seen this
type of failure before when I've specified an optimization, crossbuilding
i386 under FreeBSD.  For some reason, I'm doing this again, specifying

#   compile  GENERIC/icp.o
elf-gcc -ffreestanding -pipe -O1 -Werror -Wall -Wno-main -Wno-format-zero-length
 -Wpointer-arith -Wmissing-prototypes -Wstrict-prototypes -Wno-sign-compare -fno
-zero-initialized-in-bss -Di386 -I. -I/stand/NetBSD-src/src/sys/arch -I/stand/Ne
tBSD-src/src/sys -nostdinc -DLKM -DMAXUSERS=32 -D_KERNEL -D_KERNEL_OPT -I/stand/
NetBSD-src/src/sys/dist/ipf -c /stand/NetBSD-src/src/sys/dev/ic/icp.c
/stand/NetBSD-src/src/sys/dev/ic/icp.c: In function `icp_init':
/stand/NetBSD-src/src/sys/dev/ic/icp.c:143: warning: `i' might be used uninitial
ized in this function

I suspect this is a problem because in icp.c lines 174 and following,
one can goto bail_out, while `i' is first used in line 204.  The second
line of bail_out: at line 430 compares against `i'.

Note that this is not what was backed out in rev 1.12 of this file.

(Strange, now I *really* wonder how I was building without forcing -O1
to get successful builds recently...)

I believe a couple other files suffer this problem too when building
GENERIC, as I've reported (with patches) in the past.  I'll try again,
but my build machine is horribly slow.

barry bouwsma