[FGS] Location of support libs and headers

Daniel Morissette dmorissette at dmsolutions.ca
Tue Jul 27 11:13:07 EDT 2004


Frank,

How have you been dealing with system lib and header dependencies until
now in your FGS builds?

For instance, GDAL pulls libjpeg.so.3, but we don't see any traces of
the jpeg source anywhere in the FGS trees. I have to assume that GDAL
was built against the system's libjpeg-dev that came in /usr/lib and
/usr/include, and that libjpeg.so.3 was manually copied to the FGS base
package after.

Is this what you did?

Should we not keep a copy of all the support libs that we use under the
FGS tree, perhaps in a FGS source tree, and get rid of all development
libs and includes in the /usr/lib and /usr/include system directories? I
fear that if someone upgrades a lib in the system directories then we
will end up compiling new packages against a different versions of
header files than what was used for older packages before the upgrade,
which will potentially lead to runtime crashes.

We also noticed another problem on this system with libpng: there are
libpng12.so and libpng.so under /usr/lib, but only one png.h in
/usr/include. Your GDAL 1.2.0 linked with libpng.so, but GD grabs
libpng12.so by default, so we've ended up with duplicate copies of
libpng in our MapServer binaries. Both libpng.so and libpng12.so are
pointing to binary files of the same size but that differ according to
'diff'. It's hard to tell whether they are really different or not and
if it's safe to interchange them. I think it would be simpler/safer to
drop them from the system lib and include dirs, and only work with our
own libpng locally.

I think we'll find more of those kind of problems as we go, so let's see
if there really is a problem or if I just misunderstood the way you've
been doing your builds and we'll go from there.

Daniel
--
------------------------------------------------------------
  Daniel Morissette               dmorissette at dmsolutions.ca
  DM Solutions Group              http://www.dmsolutions.ca/
------------------------------------------------------------





More information about the Foss-gis-suite mailing list