On 2011-11-23 19:27-0800 Alan W. Irwin wrote:
> Here are my further plans:
>
> Implement the NON_TRANSITIVE option for our build system that will
> drop transitive linking for the CMake case and replace Requires: by
> Requires.private: for the pkg-config case. This option will be
> experimental and will default to OFF. Then I plan to go through our
> languages (starting with C++) one by one to deal with all the various
> issues that are found with turning that option ON.
>
DONE and completely tested for the shared library case on my Debian
stable platform for C++ (as of revision 12033). The "ldd -u" results
for the C++ examples are extremely abbreviated compared to those for
C, e.g.,
software@raven> ldd -u examples/c/x01c
Unused direct dependencies:
/home/software/plplot_svn/HEAD/build_dir/src/libplplotd.so.11
/lib/libm.so.6
/usr/lib/libltdl.so.7
/lib/libdl.so.2
/home/software/plplot_svn/HEAD/build_dir/lib/csa/libcsirocsa.so.0
/home/software/plplot_svn/HEAD/build_dir/lib/nn/libcsironn.so.0
/home/software/qhull/install/lib/libqhull.so.5
/home/software/plplot_svn/HEAD/build_dir/lib/qsastime/libqsastime.so.0
/usr/lib/libfreetype.so.6
software@raven> ldd -u examples/c++/x01
Unused direct dependencies:
/home/software/plplot_svn/HEAD/build_dir/bindings/c++/libplplotcxxd.so.10
/lib/libm.so.6
/lib/libgcc_s.so.1
In fact, the rpmlint tests for unused direct dependencies may not be
quite equivalent to "ldd -u" since that command obviously provides
erroneous results above, e.g., we know that direct linking of
libplplotcxxd and libm is required for the C++ examples and those
libraries are definitely used as well contrary to what is reported by
"ldd -u". A similar situation may apply for libgcc_s so that you
might get a completely clean rpmlint report for the C++ examples. (But
currently libplplotcxx will still have warning messages because I
haven't yet implemented non-transitive linking for our C libraries).
Another issue is my Debian stable system had no problems with the
call to c_plrandd (resolved only by libplplotd) in C++ example 17 even
though I checked that in all the three cases (build tree and
traditional and CMake-based installed examples), the example was
linked just with libplplotcxxd and not libplplotd.
So the only way I can explain this difference between my results and
yours (assuming you confirm problems with example 17 still persist) is
the Debian stable run-time loader resolves symbols by promiscuously
searching through all the indirectly linked libraries (e.g.,
libplplotd linked by libplplotcxxd but not the C++ examples) while the
Fedora run-time loader only searches through directly linked libraries
(i.e., libplplotcxxd for the C++ examples) so it cannot resolve the
c_plrandd symbol in x17.cc.
Orion, just for now I am going to leave the call to plrandd (a.k.a.
c_plrandd) in examples/c++/x17.cc as a test to see whether you continue
to get the bad C++ results on Fedora because of that.
Here is how I tested the C++ case (and I suggest you do the same)
where PL_PREFIX is an environment variable set to your preferred
PLplot install prefix which should be initially empty just like the
directories I create below with mkdir.
mkdir build_dir
cd build_dir
cmake -DCMAKE_INSTALL_PREFIX=$PL_PREFIX \
-DBUILD_TEST=ON -DDEFAULT_NO_BINDINGS=ON -DENABLE_cxx=ON \
-DNON_TRANSITIVE=ON /path/to/top/of/source_tree
# Build tree test:
make test_diff_psc
# Prerequisite for installed examples tests:
make install
# CMake-based installed examples:
mkdir install_build_dir
cd install_build_dir
cmake $PL_PREFIX/share/plplot5.9.9/examples
make test_diff_psc
# Traditional installed examples:
cd ..
mkdir traditional_build_tree
cp -a $PL_PREFIX/share/plplot5.9.9/examples traditional_build_tree
cd traditional_build_tree/examples
make compare
If C++ example 17 gives trouble (which I assume it will) please
replace the plrandd C call with the correct C++ form and try the above
tests again. If that works, I will commit the change as well.
Assuming you do prove that Fedora is more unforgiving than Debian
stable, then my plan is to get rid of all errors generated by
-DNON_TRANSITIVE=ON one language at a time (just like I have just done
for C++), and then pass off the further testing of that language to
you to see what further changes have to be made for that language to
satisfy Fedora. I am game to keep going with this as long as you
are!
Alan
__________________________
Alan W. Irwin
Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).
Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________
Linux-powered Science
__________________________
------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure
contains a definitive record of customers, application performance,
security threats, fraudulent activity, and more. Splunk takes this
data and makes sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-novd2d
_______________________________________________
Plplot-devel mailing list
Plplot-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/plplot-devel
