This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
Re: Compilation of GSL 0.7 (CVS) on Visual C++ 6.0 mostly done ...
Hello again,
After the last e-mail from Brian Gough I have checked out the
CVS version of gsl (Apr. 2nd, 2001) and I have started again with it
on Visual C++. Now I have compiled gsl on Visual C++ with all
the test (some of them dosen't links) passed, althougth I have to
look into it deeper (log the results and see).
In http://www.isys.dia.fi.upm.es/~jmbuena you can find:
- compiling_libgsl_output.txt
- compiling_tests_output.txt
- compiling_libgslcblas_output.txt
with the output of Visual C++ 6.0 when compiling GSL.
And:
- gsl-0.7-20010402.zip
with all the gsl with the changes for Visual C++. The CVS dirs are
included so, it will not be difficult to merge the changes (if they're
right, of course).
This is a quick summary of what I have now:
====================
INSTALL
====================
- All the Visual C++ 6.0 project files and related stuff is in
msvc directory.
- The intention is to have three dirs (projects) on msvc:
-libgsl
-libcblas (or something similar - I have to look into this ;-) -)
-tests (testing ptrograms)
- I'm following the steps on the Makefile.am files over the gsl dirs,
trying to compile everything in the same order.
- First is needed to have soft links on gsl dir to all the gsl*.h files
over all the tree. On windows is needed to copy all theese files to
gsl. There is a msdos bach file on msvs to do this:
copy_gsl_headers.bat
- The config.h is not generated (there is not autoconf/automake) so
I have generated it by cygwin and I'm modifying it throught try and
error. The definitive version of config.h will be in msvc/config.h.vc
and the user will have to copy it to the gsl root directory.
- The same is for gsl_version.h is not generated and the VC++ version is
in msvc/gsl_version.h.vc.
====================
GSL LIBRARIES
====================
- There is some missing functions in the math.h of Visual C++ 6.0:
acosh, asinh, atanh, hypot.
- There is a lot of warnings around const arguments to functions. In
the definition there is a const and in the declaration there is'nt and
viceversa.
- Some changes to introduce #ifdef HAVE_INLINE/#if !defined(HAVE_INLINE)
to avoid errors: "This function already has a body".
- Introduced a change into gsl_math.h to have the correct definition for
GSL_NAN, GSL_PINF and so on.
- Changed vector/view.c to include view_source.c and not init_source.c.
====================
TEST PROGRAMS:
====================
- The test programs for vector and matrix dosen't link due to
undefined symbols in libgslcblas (thouse functions are defined out on
CVS :-( ).
- The test program for linalg fail to link due to undefined symbols in
libgslcblas.
- The test program for specfun needs the same stack increase than DJGPP
compiler, so it is done in its VC++ project.
- The rng/test.c program raises an "Illegal initialitation" on the
const gsl_rng_type * rngs[] = { ..... };
In the main function. So it has been changed.
- test_fn.c: line 127: "Illegal Initialitation":
gsl_multifit_function_fdf f = { &brown_f, &brown_df, &brown_fdf, n, p, 0 };
- The test program for multifit fail to link due to undefined symbols in
libgslcblas (thouse functions are defined out on CVS :-( ).
- siman/siman_tsp.c needs to include memory.h for memcpy.
- integration/test.c: lines 78, 104, 129, 154, 179, 204, 233, 258, 283, ...
: "Illegal Initialitation".
- interpolation/test.c: lines 147, 148, ...: "Illegal Initialitation".
- ode-initval test program dosen't link due to libgslcblas missing functions.
- monte/test.c: If "Disable language extensions" (for ANSI C compliant) then on
line 82: "Illegal Initialitation" on:
gsl_monte_function Fc = { &fconst, 0, 0 };
gsl_monte_function F0 = { &f0, 0, &a };
gsl_monte_function F1 = { &f1, 0, &a };
gsl_monte_function F2 = { &f2, 0, &a };
gsl_monte_function F3 = { &f3, 0, &c };
- multiroots test program dosen't link due to libgslcblas missing
functions.
- multimin test program dosen't link due to libgslcblas missing
functions.
==============================================================
That's all by now. Don't hesiate to contact me for further
information/testing. Please, again, replay to me directly because I'm
not in the list.
Best Regards.
--
-------------------------------------------------------------------
Jose Miguel Buenaposada Biencinto
Facultad de Informática | Computer Science School
Universidad Politécnica de Madrid (UPM) | Madrid Tech. University.
Campus de Montegancedo s/n
28660 Boadilla del Monte, Madrid (ESPAÑA)| (SPAIN)
Voice +34 91 336 69 47
e-mail: jmbuena@dia.fi.upm.es
web : http://www.dia.fi.upm.es/~jmbuena
-------------------------------------------------------------------
Yo soy Precario FPU, ¿ y tu ?: http://www.precarios.org