🚀 go-pugleaf

On Mar 7, 2013, at 16:47, Dave Huang <khym@azeotrope.org> wrote:
> Should alpha be added to the package's NOT_FOR_PLATFORM? BTW, I don't
> see any m68k support either, but I haven't tried to build it on a 68k

Or if there's a way to disable the build of Boost.Context on platforms
where it's not supported, that'd be even better� I don't know much
about boost; I just want the package because it's a dependency of some
other package (which works with an older version of boost, so I assume
it doesn't use Boost.Context).

-- 
Name: Dave Huang         |  Mammal, mammal / their names are called /
INet: khym@azeotrope.org |  they raise a paw / the bat, the cat /
FurryMUCK: Dahan         |  dolphin and dog / koala bear and hog -- TMBG
Dahan: Hani G Y+C 37 Y++ L+++ W- C++ T++ A+ E+ S++ V++ F- Q+++ P+ B+ PA+ PL++

On Mar 7, 2013, at 17:03, Tim Larson <Tim.Larson@proxibid.com> wrote:

> > Or if there's a way to disable the build of Boost.Context on
> > platforms where it's not supported, that'd be even better� I don't
> > know much about boost; I just want the package because it's a
> > dependency of some other package (which works with an older version
> > of boost, so I assume it doesn't use Boost.Context).
> 
> Pretty sure I've seen platform-conditional patches in pkgsrc. Patch the filename in question out of the Makefile so it isn't built, only on those platforms.

While I suspect that's technically possible, I was thinking more along
the lines of an option to a configure script, e.g., configure
--disable-context. Except boost doesn't seem to have a configure
script. If boost is designed so that the Context stuff is an integral
part, trying to hack it back out is probably more effort than is worth
doing. However, if they've included an easy knob to turn to disable the
Context stuff, that'd be the way to go.

-- 
Name: Dave Huang         |  Mammal, mammal / their names are called /
INet: khym@azeotrope.org |  they raise a paw / the bat, the cat /
FurryMUCK: Dahan         |  dolphin and dog / koala bear and hog -- TMBG
Dahan: Hani G Y+C 37 Y++ L+++ W- C++ T++ A+ E+ S++ V++ F- Q+++ P+ B+ PA+ PL++

On Thu, Mar 07, 2013 at 04:47:12PM -0600, Dave Huang wrote:
> As far as I can tell, boost 1.53 won't build on alpha due to
> Boost.Context requiring a bit of assembly code. pkgsrc has a patch to
> add the necessary bits for sparc, but there's nothing for alpha.
> Should alpha be added to the package's NOT_FOR_PLATFORM? BTW, I don't
> see any m68k support either, but I haven't tried to build it on a 68k
> :)

I added the sparc stuff and worked with upstream to get it integrated
there and I also have alpha and m68k machines available for testing.

For alpha it sounds reasonable to add support, but is anyone actually
going to use a monster like boost-libs on m68k?

I remember how painfull it was on sparc (and I kinda cheated there with
a sparc64 machine running 32bit).

To put it blunt: isn't it much more sensible to mark boost-libs as
NOT_FOR_.... instead? Does anyone know if it fully builds w/o context?

No pun intended, if anyone realy uses it, I might even be talked into
trying to add support for the platform. Done it once, understand the
interface, know how to test...

Martin

On Mar 8, 2013, at 1:52, Martin Husemann <martin@duskware.de> wrote:

> For alpha it sounds reasonable to add support, but is anyone actually
> going to use a monster like boost-libs on m68k?

Yeah, boost is huge and takes a lot of RAM to compile; I wonder if 
there are any m68k machines that can even build it. I don't have any 
need for it on m68k, and have no objections to marking it NOT_FOR m68k. 

I'd like it to work on alpha though (without Context support is fine 
with me, if boost has an option for that�)
-- 
Name: Dave Huang         |  Mammal, mammal / their names are called /
INet: khym@azeotrope.org |  they raise a paw / the bat, the cat /
FurryMUCK: Dahan         |  dolphin and dog / koala bear and hog -- TMBG
Dahan: Hani G Y+C 37 Y++ L+++ W- C++ T++ A+ E+ S++ V++ F- Q+++ P+ B+ PA+ PL++

On Fri, Mar 08, 2013 at 01:59:36PM +0000, Tim Larson wrote:
> Sounds like a good question for upstream. Anyone here follow boost
> development?

I'll ask there (and may even try to build on mac68k just for fun, it
just takes machine time, not mine).

Martin

Looking at the boost configure stuff, this (untested) patch might do it:

Index: Makefile
===================================================================
RCS file: /cvsroot/pkgsrc/devel/boost-libs/Makefile,v
retrieving revision 1.24
diff -u -r1.24 Makefile
--- Makefile	31 Oct 2012 11:16:42 -0000	1.24
+++ Makefile	9 Mar 2013 14:29:55 -0000
@@ -10,8 +10,11 @@
 INSTALLATION_DIRS+=	lib
 
 BJAM_ARGS+=		--without-python
+
 .if !empty(OPSYS:MNetBSD) && !empty(MACHINE_ARCH:Msparc64)
 BJAM_ARGS+=		pch=off
+.elif !empty(MACHINE_ARCH:Malpha) || !empty(MACHINE_ARCH:Mm68k)
+BJAM_ARGS+=		--without-libraries=context
 .endif
 
 UNLIMIT_RESOURCES+=	stacksize


Anyone want to give it a try and report?

Thanks,

Martin

On Sat, Mar 09, 2013 at 03:30:44PM +0100, Martin Husemann wrote:
> +.elif !empty(MACHINE_ARCH:Malpha) || !empty(MACHINE_ARCH:Mm68k)
> +BJAM_ARGS+=		--without-libraries=context

If this does not work, it might be

BJAM_ARGS+=		--without-context

Martin

On Mar 9, 2013, at 8:34, Martin Husemann <martin@duskware.de> wrote:

> On Sat, Mar 09, 2013 at 03:30:44PM +0100, Martin Husemann wrote:
>> +.elif !empty(MACHINE_ARCH:Malpha) || !empty(MACHINE_ARCH:Mm68k)
>> +BJAM_ARGS+=		--without-libraries=context
> 
> If this does not work, it might be
> 
> BJAM_ARGS+=		--without-context

Yes, it's --without-context. My mac68k system is currently out of
commission, but the build works on alpha with that patch. However, the
package installation fails:

=> Creating binary package /usr/pkgsrc/packages/All/boost-libs-1.53.0.tgz
pkg_create: can't stat `/usr/pkgsrc/devel/boost-libs/work.alpha/.destdir/usr/pkg/lib/libboost_context.a'
pkg_create: can't stat `/usr/pkgsrc/devel/boost-libs/work.alpha/.destdir/usr/pkg/lib/libboost_context.so'
pkg_create: can't stat `/usr/pkgsrc/devel/boost-libs/work.alpha/.destdir/usr/pkg/lib/libboost_context.so.1.53.0'
pkg_create: lstat failed for file lib/libboost_context.a: No such file or directory
*** Error code 2

--x+6KMIRAuhnl3hBn
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: 8bit

On Sun, Mar 10, 2013 at 09:33:16AM -0500, Dave Huang wrote:
> Yes, it's --without-context. My mac68k system is currently out of
> commission, but the build works on alpha with that patch. However, the
> package installation fails:
> 
> => Creating binary package /usr/pkgsrc/packages/All/boost-libs-1.53.0.tgz
> pkg_create: can't stat `/usr/pkgsrc/devel/boost-libs/work.alpha/.destdir/usr/pkg/lib/libboost_context.a'
> pkg_create: can't stat `/usr/pkgsrc/devel/boost-libs/work.alpha/.destdir/usr/pkg/lib/libboost_context.so'
> pkg_create: can't stat `/usr/pkgsrc/devel/boost-libs/work.alpha/.destdir/usr/pkg/lib/libboost_context.so.1.53.0'
> pkg_create: lstat failed for file lib/libboost_context.a: No such file or directory

Yeah, those need to be made conditional in the PLIST.

Could you try the attached patch?

Thanks,

Martin

--x+6KMIRAuhnl3hBn
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename=patch
Content-Transfer-Encoding: 8bit

Index: Makefile
===================================================================
RCS file: /cvsroot/pkgsrc/devel/boost-libs/Makefile,v
retrieving revision 1.24
diff -u -r1.24 Makefile
--- Makefile	31 Oct 2012 11:16:42 -0000	1.24
+++ Makefile	10 Mar 2013 20:36:10 -0000
@@ -10,8 +10,16 @@
 INSTALLATION_DIRS+=	lib
 
 BJAM_ARGS+=		--without-python
+PLIST_VARS+=		context
+
 .if !empty(OPSYS:MNetBSD) && !empty(MACHINE_ARCH:Msparc64)
 BJAM_ARGS+=		pch=off
+.elif !empty(MACHINE_ARCH:Malpha) || !empty(MACHINE_ARCH:Mm68k)
+BJAM_ARGS+=		--without-context
+.endif
+
+.if empty(BJAM_ARGS:M--without-context)
+PLIST.context=		yes
 .endif
 
 UNLIMIT_RESOURCES+=	stacksize
Index: PLIST
===================================================================
RCS file: /cvsroot/pkgsrc/devel/boost-libs/PLIST,v
retrieving revision 1.18
diff -u -r1.18 PLIST
--- PLIST	29 Nov 2012 21:50:24 -0000	1.18
+++ PLIST	10 Mar 2013 20:36:10 -0000
@@ -2,9 +2,9 @@
 lib/libboost_chrono.a
 lib/libboost_chrono.so
 lib/libboost_chrono.so.${BOOST_VERSION}
-lib/libboost_context.a
-lib/libboost_context.so
-lib/libboost_context.so.${BOOST_VERSION}
+${PLIST.context}lib/libboost_context.a
+${PLIST.context}lib/libboost_context.so
+${PLIST.context}lib/libboost_context.so.${BOOST_VERSION}
 lib/libboost_date_time.a
 lib/libboost_date_time.so
 lib/libboost_date_time.so.${BOOST_VERSION}

--x+6KMIRAuhnl3hBn--

On Sun, Mar 10, 2013 at 09:40:41PM +0100, Martin Husemann wrote:
> Yeah, those need to be made conditional in the PLIST.
> 
> Could you try the attached patch?

It works (on alpha). Thanks! :)
-- 
Name: Dave Huang         |  Mammal, mammal / their names are called /
INet: khym@azeotrope.org |  they raise a paw / the bat, the cat /
FurryMUCK: Dahan         |  dolphin and dog / koala bear and hog -- TMBG
Dahan: Hani G Y+C 37 Y++ L+++ W- C++ T++ A+ E+ S++ V++ F- Q+++ P+ B+ PA+ PL++

On Mon, Mar 11, 2013 at 10:24:29PM -0500, Dave Huang wrote:
> It works (on alpha). Thanks! :)

Thanks for testing! I will commit it.

Martin

>>> Could you try the attached patch?
>> 
>> It works (on alpha). Thanks! :)
>
> It might be noteworthy that attempting to compile boost-libs on a netbsd-6 
> mac68k system with 68 megs caused it to panic and reboot. I'm testing it on a 
> 256 meg Amiga now.

It compiles and installs fine on m68k (with memory), but I haven't tested 
it with anything yet.

John