AMBER Archive (2007)

Subject: Re: [Fwd: Re: AMBER: xleap patch for xorg-7.3 ??]

From: M. L. Dodson (activesitedynamics_at_comcast.net)
Date: Thu Nov 15 2007 - 11:28:28 CST

Xuebin Qiao wrote:
> >If I may make a suggestion, rather than hack the configure script,
> >why not limit the fix to the makefile in the patch? Something along
> >the line of:
>
> It is unwise to hack in makefile using non-portable features. Put all
> ugly things into a single place would be more decent design.
>

Yes, /bin/sh certainly has more, and more portable, features to
implement this "ugly thing". :-)

The only thing is that linking to libXau and libXdmcp should
probably occur right before -lm -lpthread, as linking to the libs
earlier on the command line probably causes the dependencies on
libXau and libXdmcp. If this is so the link command line would
probably have to be set up in the configure script, not in the
makefile. I.e., it should be in config.h after configuration.

Maybe we should take this off list as most people could probably
care less.

> >The only corner case I can think of that this might miss is a
> >partial or messed up X installation.
>
> maybe yes or maybe not, function dlsym is used when programmer need to
> load certain dynamic library by hand, and it resides in libc.so. So I
> guess add -lc may be ok, though some other unresolved symbols may appear
> too. :-) If only this symbol missed, it would has nothing to with X, but
> may due to a unsuccessful partial upgrading to the OS.
>
> qxb

Correct. I was not even considering libc, only corner cases
having to do with the need or no need to link to libXau and
libXdmcp. It would be a fairly strange contemporary c compiler to
not be able to link to its libc in some automatic fashion, though,
would it not?

Bud Dodson

>
> On Nov 15, 2007 2:51 AM, M. L. Dodson <activesitedynamics_at_comcast.net
> <mailto:activesitedynamics_at_comcast.net>> wrote:
>
> I sent this out last week, but it never appeared on the list.
> So here we go again.
>
> Bud Dodson
>
> -------- Original Message --------
> Subject: Re: AMBER: xleap patch for xorg-7.3 ??
> Date: Fri, 09 Nov 2007 10:39:15 -0600
> From: M. L. Dodson <activesitedynamics_at_comcast.net
> <mailto:activesitedynamics_at_comcast.net>>
> Reply-To: activesitedynamics_at_comcast.net
> <mailto:activesitedynamics_at_comcast.net>
> To: amber_at_scripps.edu <mailto:amber_at_scripps.edu>
>
> Xuebin Qiao wrote:
> > Hi,
> >
> > The Xorg7 has been designed to a new modularized architecture
> instead of
> > the old monolithic source tree. So it may break many old
packages,
> > especially for those are linked with hand-coded linker options.
> >
> > qxb
> >
> > On Nov 9, 2007 10:03 AM, David A. Case <case_at_scripps.edu
> <mailto:case_at_scripps.edu>
> > <mailto:case_at_scripps.edu <mailto:case_at_scripps.edu>>> wrote:
> >
> > On Thu, Nov 08, 2007, M. L. Dodson wrote:
> > >
> > > I found the following patch to be necessary for
successfully
> > > linking xaLeap with xorg-7.3
> > >
> >
> > > --- leap/src/leap/Makefile.orig 2007-10-03 09:59:
> > 50.000000000 -0500
> > > +++ leap/src/leap/Makefile 2007-11-08
> 12:08:59.000000000 -0600
> > > @@ -114,7 +114,7 @@
> > > # XlcDL.o(.text+0xe2): undefined reference to `dlsym'
> > > #
> > > XALEAP_LIB = ../Xraw/libXaw.a ../Wc/libWcLeap.a
> ../Xpm/libXpm.a \
> > > - ../Xmu/libXmu.a $(XLIBS) -lXt -lXext -lSM -lICE
> -lX11 $(LM)
> > -lpthread
> > > + ../Xmu/libXmu.a $(XLIBS) -lXt -lXext -lSM -lICE
> -lX11 -lXau
> > -lXdmcp $(LM) -lpthread
> > >
> > > OBJ = $(OA) $(OG) $(OO) $(OL)
> > >
> >
> > Thanks for the report. But some more information might help:
> >
> > What is the error when you don't have -lXau -lXdmcp specified?
> Do you
> > undertstand why xorg-7.3 needs these but not other
> X-implementations?
> >
>
> Unresolved symbols in the link. No, although it may have
> something to do with factoring the functionality of X,
> contemporaneously at least, with the build factorization qxb
> described above.
>
> > (extra credit:) how standard are libXau and libXdmcp? They
> seem to
> > be present
> > on IRIX, Cygwin and Linuces, but I don't know about AIX or
MacOS
> > (for example).
> >
>
> I can speak only for recent FreeBSD X installs and MacOS 10.4.10
> (my laptop). On the mac, xdpyinfo says it is XFree86 4.4.0, and
> it is at least two years since I installed it (several OS minor
> version upgrades since then). I.e., it predates the xorg split
> from XFree86, I think. Both libs present on these OS types.
>
> > These "standard X" libraries are ancient...why would xorg move
> > things around
> > in a way almost guaranteed to break old programs?
> >
>
> The ABI is probably still intact, but the implementations may now
> call functions in the new libs. This may be associated with the
> functionality factorization I implied above.
>
> > I'm happy to make the above patch if it is safe for all
machines.
> > Otherwise,
> > we'll have to add code to the configure script to
> special-case this
> > out.
> >
> > ...dac
> >
>
> If I may make a suggestion, rather than hack the configure script,
> why not limit the fix to the makefile in the patch? Something along
> the line of:
>
> XALEAP_LIB = ../Xraw/libXaw.a ../Wc/libWcLeap.a ../Xpm/libXpm.a \
> ../Xmu/libXmu.a $(XLIBS) -lXt -lXext -lSM -lICE -lX11
> <test for the existence of the libs>
> if (libs present)
> XALEAP_LIB += -lXau -lXdmcp $(LM) -lpthread
> else
> XALEAP_LIB += $(LM) -lpthread
>
> My SysV make-foo is not up to this test (I'm a Berkeley make kind
> of guy), but this should be possible, at least for gmake, I would
> think.
>
> The only corner case I can think of that this might miss is a
> partial or messed up X installation. It may give a misleading
> error message in that case.
>
> Bud Dodson
> 

-- 
M. L. Dodson
Business Email: activesitedynamics-at-comcast-dot-net
Personal Email:	mldodson-at-comcast-dot-net
Phone:	eight_three_two-56_three-386_one
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu