Re: [changeset] GraphicsMagick++ configuration

[Thomas Weber]

On Sun, Aug 16, 2009 at 11:09:09PM +0200, Thomas Weber wrote:
> On Thu, Aug 06, 2009 at 04:15:56PM -0400, John W. Eaton wrote:
> > On  6-Aug-2009, David Grundberg wrote:
> > 
> > | John W. Eaton skrev:
> > | > On  6-Aug-2009, David Grundberg wrote:
> > | >
> > | > | I've rearranged the GraphicsMagick++ configuration. I had some 
> > trouble 
> > | > | since I'm running a custom GraphicsMagick installation. The Octave 
> > build 
> > | > | system was running GraphicsMagick++-config during make. It was 
> > missing 
> > | > | ldflags and using only the basename of the config executable (as 
> > opposed 
> > | > | to a full path).
> > | > | 
> > | > | I've changed it so that GraphicsMagick++-config is only run in the 
> > | > | configure script. Also introduced MAGICK_CONFIG as a precious 
> > variable.
> > | >
> > | > I removed --ldflags to fix the following mysterious problem on my
> > | > system:
> > | >
> > | >   
> > https://www-old.cae.wisc.edu/pipermail/octave-maintainers/2009-July/012879.html
> > | >
> > | > So I don't think I can put it back without breaking __magick_read__
> > | > again, at least for me and other Debian users.
> > | >
> > | > What are -Wl,-z,relro and -pie doing in the ldflags anyway?  Are those
> > | > arguments not present on your system?  Maybe they are present on mine
> > | > because of the way Debian builds the graphics magick library package?
> > | > Perhaps these flags make sense for creating the graphics magick
> > | > library itself, but I can't see any reason for them to be required to
> > | > link the graphics magick library with __magick_read__.oct.
> > | >
> > | > jwe
> > | >   
> > | So that's why --ldflags was taken away! I don't have that problem. This 
> > | is the output from my config (where I'm building):
> > | 
> > | $ 
> > | 
> > octave-patching/dependencies/graphicsmagick-install/bin/GraphicsMagick++-config
> >  
> > | --ldflags --libs
> > | 
> > -L/Home/staff/davidg/octave-patching/dependencies/graphicsmagick-install/lib
> >  
> > | -L/usr/lib -L/usr/lib
> > | -lGraphicsMagick++ -lGraphicsMagick -llcms -ltiff -lfreetype -ljasper 
> > | -ljpeg -lpng -lwmflite -lXext -lSM -lICE -lX11 -lbz2 -lxml2 -lz -lm 
> > | -lgomp -lpthread
> > | 
> > | My ubuntu 9.04 box says this about the managed package (haven't tried 
> > | building against it):
> > | address@hidden:~$ GraphicsMagick++-config --ldflags --libs
> > | -L/usr/lib -Wl,-Bsymbolic-functions -L/usr/lib/X11 -L/usr/lib -L/usr/lib
> > | -lGraphicsMagick++ -lGraphicsMagick -llcms -ltiff -lfreetype -ljasper 
> > | -ljpeg -lpng -lwmflite -lXext -lSM -lICE -lX11 -lbz2 -lxml2 -lz -lm 
> > | -lpthread
> > 
> > So what should we do about this?  
> 
> These are different versions. The Ubuntu version is 1.1.11-something,
> your Debian versions is 1.3.5-something.
> 
> > around it by filtering out everything except -L options from the
> > --ldflags output, but I'd rather see it fixed in graphics magick.
> > What should graphics magick really be storing in the config script
> > for --ldflags?  Should options like -pie and -Wl,-z,relro really
> > appear there?  Or is this just a Debian packaging problem?
> 
> The Debian packaging adds the -pie and -Wl options when building
> GraphicsMagick to the LDFLAGS. GraphicsMagick then copies LDFLAGS into its
> --ldflags output. Looking at Ubuntu's build logs, the same is already
> happening with newer versions of graphicsmagick.
> 
> I'll ask the Debian maintainer of GraphicsMagick about it.

Okay, his answer is below. Short version: we should go we with
pkg-config.

============================================================================
Historically, the output of the *Magick*config scripts has been documented as
the list of compiler/linker options *Magick was built with. This makes sense
when build additional modules, but it's rather nonsensical for anything that
just wants to link the C or C++ bindings. I recommend to use pkg-config for
these kinds of applications instead. The hardening options aren't included
there.
============================================================================

        Thomas