This is the mail archive of the gsl-discuss@sourceware.cygnus.com mailing list for the GSL project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: broken after libtoolswitch on


It seems complicated....

noinst_LTLIBRARIES (the current setting in libtoolswitch.pl) makes a
single libgsl. This has the side effect of removing the intermediate
files in each directory. So you have to build the whole library to get
anything.

Using just LTLIBRARIES gives an error from automake,

  Makefile.am:13: `LTLIBRARIES' is an anachronism

The alternative is pkglib_LTLIBRARIES which does build and install
each of the libraries separately. But this has the side effect that if
you try to build a single libgsl there is an error,

  libtool: link: error: cannot link shared libraries into libtool libraries

So it does not seem possible to build a single libgsl and also build
separate shared libraries at the same time.

Thomas Walter writes:
 > 
 > Hello,
 > now after switching to Debian/Linux I did an 'cvs update' of my gsl
 > tree, copied it to another fresh location and started:
 >       libtoolize -c -f
 >       libtoolswitch.pl on
 >       autogen.sh
 >       make
 > 
 > After some time I looked what it is doing and I am surprised.
 > Libtool generates the static and the shared libraries in each
 > subdirectory and then it removes '.libs/lib*.la'.
 > A check with 'cd sum; make install -n' show it wants to install the
 > headers only 8-(.
 > I located the problem (I think) in the script 'libtoolswitch.pl'.
 > It has the rule
 >    s/pkglib_LIBRARIES/noinst_LTLIBRARIES/g ;
 > which says not to install the library.
 > I think it should be
 >   s/pkglib_LIBRARIES/LTLIBRARIES/g ;
 > 
 > Can anybody verify this?
 > 
 > A side question:
 > Does anybody know an algorithm to solve a five-diagonal system?
 > This could speed up my implementation of an approximating spline.
 > Currently I use a LU-decomposition (Crout) without pivoting.  This
 > works well because the matrix is diagonal dominant.
 > Hints are welcome.
 > 
 > Bye
 > Thomas
 > 
 > -- 
 > SUSHI?  Nein, Danke.  Ich esse keine Kder.
 > 
 > ----------------------------------------------
 > Dipl. Phys. Thomas Walter
 > Inst. f. Physiklische Chemie II
 > Egerlandstr. 3				Tel.: ++9131-85 27326 / 27330
 > 91058 Erlangen, Germany			email: walter@pctc.chemie.uni-erlangen.de

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]