Dane, Ludwig:

Once again, thanks again for your help.  I'm proud to say I was finally able
to get mapnik installed! :-)

I ended up making a number of adjustments to get it installed and working on
one of my servers, which is a Plesk-controlled RHEL 5 environment.  Here
were the steps I ended up taking:

   - Modified /etc/ld.so.conf so that the /usr/local/lib directory was
   listed FIRST.  I don't totally understand how ld interacts with the gcc
   compiler, but the ldconfig -v output would pick up /usr/lib modules prior to
   finding the /usr/local/lib modules.
   - Updated the 'LIBS' PathVariable references in SConstruct for the icu,
   png, jpeg, tiff, and proj requirements.  I had to remove the references to
   the LIBDIR_SCHEMA variable and point the PathVariable at the install
   directory for each of the files.  Otherwise it was questionable if the
   configure step would find those header files.
   - Comment out the reference to the shapeindex plugin. I never could get
   it to compile.

I ended up abandoning my efforts to get the 0.6.1 branch installed, so these
steps were taken against the 0.7 branch.  These three adjustments allowed me
to install mapnik and generate a successful test via Python.  Now on the fun
business of moving mapnik into production.

Thanks again for the help.  It's most appreciated.

Thomas

On Wed, Jan 6, 2010 at 1:29 PM, Dane Springmeyer <[email protected]> wrote:

>
> On Jan 6, 2010, at 10:04 AM, Thomas Woodham wrote:
>
>  Sorry, forgot to include that detail.  That boost version (1.33) is a
>> prerequisite for a number of virtual server packages, similar to Ludwig's
>> situation.  I can't get rid of it, though I thought about pointing all the
>> boost sym links to the propper versions installed in /usr/local/lib, though
>> that seems to be asking for trouble.  I may go back and reinstall boost and
>> set the prefix to /usr so it overwrites that version.
>>
>
>
> Ah, gocha. Well installing latest boost on top of /usr/lib could be bad
> news for existing applications that are using it.
>
> I think the issue is that our SCons implementation stores several default
> search paths as /usr/lib, so setting the BOOST_LIBS/INCLUDES to /usr/local
> may not secure the linking because -L/usr/lib is still put on the linkers
> path.
>
> Short of reworking some things in the build scripts to allow forcing
> /usr/local to come before /usr/ in g++/linker flags, what you can try to do
> is set the LIBS/INCLUDES for each of 'PNG', 'JPEG', 'TIFF','PROJ','ICU'
> dependencies to /usr/local (if possible) to ensure that SCons will not
> default to putting /usr/lib onto the linker.
>
> Sorry, for inelegant solution.
>
> Dane
>
>
_______________________________________________
Mapnik-users mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/mapnik-users

Reply via email to