Date: Fri, 27 Apr 2012 14:04:44 -0600 From: Warner Losh <imp@bsdimp.com> To: Mark Tinguely <marktinguely@gmail.com> Cc: arm@FreeBSD.org, Tim Kientzle <kientzle@FreeBSD.org> Subject: Re: Cross-buildworld works but not native build? Message-ID: <495386AF-37AB-450D-B78C-2A6282083587@bsdimp.com> In-Reply-To: <CAP%2BM-_Ev1TE7JEmSRbVPg3HRwMfsPVyNyR92X5UONdxmGryuaQ@mail.gmail.com> References: <9AD7075B-B85D-40DB-84B7-FD630B858A30@freebsd.org> <CAP%2BM-_Ev1TE7JEmSRbVPg3HRwMfsPVyNyR92X5UONdxmGryuaQ@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 27, 2012, at 1:51 PM, Mark Tinguely wrote: > On Fri, Apr 27, 2012 at 1:48 AM, Tim Kientzle <kientzle@freebsd.org> = wrote: >> I've been working with the projects/armv6 tree and have encountered a = very confusing situation. >>=20 >> On i386, this works: >> $ make TARGET_ARCH=3Darm TARGET_CPUTYPE=3Darmv6 buildworld >>=20 >> If I take the resulting world and run it on arm, then the following = fails (with the exact same source): >> $ make buildworld >> =85. >> cc -O -pipe -fpic -fvisibility=3Dhidden -DVISIBILITY_HIDDEN = -std=3Dgnu99 -Wsystem-headers -Werror -Wall -Wno-format-y2k = -Wno-uninitialized -Wno-pointer-sign -c = /usr/src/lib/libcompiler_rt/../../contrib/compiler-rt/lib/umoddi3.c -o = umoddi3.o >> cc -O -pipe -fpic -fvisibility=3Dhidden -DVISIBILITY_HIDDEN = -std=3Dgnu99 -Wsystem-headers -Werror -Wall -Wno-format-y2k = -Wno-uninitialized -Wno-pointer-sign -c = /usr/src/lib/libcompiler_rt/../../contrib/compiler-rt/lib/umodti3.c -o = umodti3.o >> cc -O -pipe -fpic -fvisibility=3Dhidden -DVISIBILITY_HIDDEN = -std=3Dgnu99 -Wsystem-headers -Werror -Wall -Wno-format-y2k = -Wno-uninitialized -Wno-pointer-sign -c = /usr/src/lib/libcompiler_rt/__sync_fetch_and_add_4.c -o = __sync_fetch_and_add_4.o >> In file included from = /usr/src/lib/libcompiler_rt/__sync_fetch_and_op_n.h:31, >> from = /usr/src/lib/libcompiler_rt/__sync_fetch_and_add_4.c:6: >> /usr/obj/usr/src/tmp/usr/include/machine/atomic.h: In function = 'atomic_cmpset_32': >> /usr/obj/usr/src/tmp/usr/include/machine/atomic.h:491: error: = 'ARM_RAS_START' undeclared (first use in this function) >> /usr/obj/usr/src/tmp/usr/include/machine/atomic.h:491: error: (Each = undeclared identifier is reported only once >> /usr/obj/usr/src/tmp/usr/include/machine/atomic.h:491: error: for = each function it appears in.) >> /usr/obj/usr/src/tmp/usr/include/machine/atomic.h: In function = 'atomic_add_32': >> /usr/obj/usr/src/tmp/usr/include/machine/atomic.h:516: error: = 'ARM_RAS_START' undeclared (first use in this function) >>=20 >>=20 >> Looking at the source, ARM_RAS_START really does seem to be = undeclared (it's declared in sysarch.h, but atomic.h only includes = sysarch.h for kernel builds). >>=20 >> So it looks to me like the cross-buildworld should fail also. In any = case, it's not clear why the two aren't behaving the same way. >>=20 >> Tim >>=20 >=20 > Looks like ARM_ARCH_6 or ARM_ARCH_7A is not defined. >=20 > ARM_RAS_START/ARM_RAS_END has been removed out of the the ARMv6/ARMv7 = in favor > of the ldrex/strex operations. >=20 > Also the ARM_TP_ADDRESS is not used nor mapped and uses a built-in > processor thread > register. I think it would work in the current tree if you added = TARGET_CPUTYPE=3Darmv6 to the buildworld. I'm hoping to fix issues like this in my armv6 patches.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?495386AF-37AB-450D-B78C-2A6282083587>