Date: Tue, 11 Jan 2011 14:00:07 -0500 From: John Baldwin <jhb@freebsd.org> To: Doug Barton <dougb@freebsd.org> Cc: svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org Subject: Re: svn commit: r217213 - head/lib/bind Message-ID: <201101111400.08224.jhb@freebsd.org> In-Reply-To: <4D2CA626.4030207@FreeBSD.org> References: <201101092347.p09NlB4M060802@svn.freebsd.org> <201101110847.24284.jhb@freebsd.org> <4D2CA626.4030207@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday, January 11, 2011 1:49:10 pm Doug Barton wrote:
> On 01/11/2011 05:47, John Baldwin wrote:
> > On Tuesday, January 11, 2011 12:04:56 am Doug Barton wrote:
> >> On 01/10/2011 11:33, John Baldwin wrote:
> >>> On Sunday, January 09, 2011 6:47:11 pm Doug Barton wrote:
> >>
> >>>> # Use the right version of the atomic.h file from lib/isc
> >>>> -ISC_ATOMIC_ARCH=${MACHINE_CPUARCH:S/i386/x86_32/:S/amd64/x86_32/}
> >>>> +.if ${MACHINE_ARCH} == "amd64" || ${MACHINE_ARCH} == "i386"
> >>>
> >>> This should use MACHINE_CPUARCH instead of MACHINE_ARCH?
> >>
> >> When this block was first written I don't think _CPUARCH existed at all,
> >> and/or had not been MFC'ed. Subsequently, nwhitehorn made the change in
> >> r209886 to use _CPUARCH in both places, however the diff from Warner's
> >> commit in r217071 (which is what I based my change on) clearly showed
> >> _ARCH
> >>
> >
(http://svn.freebsd.org/viewvc/base/head/lib/bind/config.mk?r1=209886&r2=217071).
> >> So, I'm not sure where svn went sideways there, but it's definitely
"odd."
> >
> > Huh? Warner's change only used MACHINE_CPUARCH:
>
> That's not what I said at all. I took a lot of time to research and
> document what I wrote, the least you could do is take a few minutes to
> read it carefully.
>
> This is the diff from the previous version to nwhitehorn's version:
>
> --- head/lib/bind/config.mk 2009/06/01 21:58:59 193280
> +++ head/lib/bind/config.mk 2010/07/10 17:46:53 209886
> @@ -45,7 +45,7 @@
> CFLAGS+= -DUSE_MD5
>
> # Endianness
> -.if ${MACHINE_ARCH} == "powerpc" || ${MACHINE_ARCH} == "sparc64"
> +.if ${MACHINE_CPUARCH} == "powerpc" || ${MACHINE_CPUARCH} == "sparc64"
> CFLAGS+= -DWORDS_BIGENDIAN
> .endif
>
> @@ -67,7 +67,7 @@
> .if ${MACHINE_ARCH} == "amd64" || ${MACHINE_ARCH} == "i386"
And _this_ line is still MACHINE_ARCH here, which is wrong. This should use
MACHINE_CPUARCH. That is the only line I have been talking about the entire
time.
> ISC_ATOMIC_ARCH= x86_32
> .else
> -ISC_ATOMIC_ARCH= ${MACHINE_ARCH}
> +ISC_ATOMIC_ARCH= ${MACHINE_CPUARCH}
> .endif
>
> # Optional features
>
>
> This is the diff from what should have been Nathan's version to Warner's:
>
> --- head/lib/bind/config.mk 2010/07/10 17:46:53 209886
> +++ head/lib/bind/config.mk 2011/01/06 21:07:51 217071
> @@ -1,6 +1,7 @@
> # $FreeBSD$
>
> .include <bsd.own.mk>
> +.include <bsd.endian.mk>
>
> # BIND version number
> .if defined(BIND_DIR) && exists(${BIND_DIR}/version)
> @@ -45,7 +46,7 @@
> CFLAGS+= -DUSE_MD5
>
> # Endianness
> -.if ${MACHINE_CPUARCH} == "powerpc" || ${MACHINE_CPUARCH} == "sparc64"
> +.if ${TARGET_ENDIANNESS} == 4321
> CFLAGS+= -DWORDS_BIGENDIAN
> .endif
>
> @@ -64,11 +65,7 @@
> .endif
>
> # Use the right version of the atomic.h file from lib/isc
> -.if ${MACHINE_ARCH} == "amd64" || ${MACHINE_ARCH} == "i386"
> -ISC_ATOMIC_ARCH= x86_32
> -.else
> -ISC_ATOMIC_ARCH= ${MACHINE_CPUARCH}
> -.endif
> +ISC_ATOMIC_ARCH=${MACHINE_CPUARCH:S/i386/x86_32/:S/amd64/x86_32/}
Note here how in the Warner's diff, he replaced one use of MACHINE_ARCH (the
amd64 and i386 comparisons) and one use of MACHINE_CPUARCH (the other cases)
with always using MACHINE_CPUARCH.
>
> # Optional features
> .if ${MK_BIND_LARGE_FILE} == "yes"
> @@ -121,4 +118,3 @@
>
> PTHREAD_DPADD= ${LIBPTHREAD}
> PTHREAD_LDADD= -lpthread
>
>
> Please note that the code Warner removed had the following line:
> -.if ${MACHINE_ARCH} == "amd64" || ${MACHINE_ARCH} == "i386"
>
> I have no idea how svn managed to fsck that up, but that is what I
> reverted back to when I made the commit in r217213. If you're still
> confused, please review the svn history yourself.
svn did the right thing. The code was broken before Warner's fix. To be
clear, this is all I think needs to change:
Index: head/lib/bind/config.mk
===================================================================
--- head/lib/bind/config.mk (revision 217272)
+++ head/lib/bind/config.mk (working copy)
@@ -65,7 +65,7 @@
.endif
# Use the right version of the atomic.h file from lib/isc
-.if ${MACHINE_ARCH} == "amd64" || ${MACHINE_ARCH} == "i386"
+.if ${MACHINE_CPUARCH} == "amd64" || ${MACHINE_CPUARCH} == "i386"
ISC_ATOMIC_ARCH= x86_32
.else
ISC_ATOMIC_ARCH= ${MACHINE_CPUARCH}
--
John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201101111400.08224.jhb>