[PURRS-devel] Re: [Fwd: fi_lib++] (2nd attempt: first time sent on 06/02/2002)

[Roberto Bagnara]

J.Wolff v. Gudenberg wrote:
> Dear Roberto Bagnara,
> 
> For what application do you need complex interval arithmetic ?
>     Do you need the extended mode ?

Dear Wolff,

our field of application is data-flow analysis and abstract interpretation.
More specifically we now need complex interval arithmetic in a project that
deals with the automatic approximate solution of difference equations.
We need the extended mode; we would also need intervals with open boundaries,
but we can live without that for the moment.

>   In the next future we are not planning to implement complex arithmetic
> for filib++.
>   but we would appreciate, if your group can add this facility for the
> library.

OK, I have just sent a message to the authors of CoStLy to see if anyone
is doing the same thing.  We will start working on that immediately if the
answer is negative.

>   We are investigating a way to use the same cvs archive behind our
> firewall.

We have already put filib++ on our CVS server so that we can easily
track our and your changes.

>>From the math viewpoint of complex arithmetic, we think that it is not
> trivial to define 
> the extended mode. We currently see no ressources on our site for
> research in this area.
> But we are positively interested in the subject.

OK.

I have just fetched the new version of filib++ (the one of February 6)
and I am having some problems with it.  Firstly, the LIBDIR directory
must be created by hand.  In fact, if one tries to do `make libs'
things end up with

ar rv /home/roberto/filib++/interval/macro/lib/fi_lib.a Interval.o  acos.o      acosh.o acot.o      acoth.o     asin.o asinh.o     atan.o      atanh.o cos.o       cosh.o      cot.o coth.o      exp.o       exp10.o exp2.o      expm1.o log.o       log10.o     log1p.o log2.o      pow.o       q_acos.o q_acot.o    q_acsh.o    q_acth.o q_asin.o    q_asnh.o    q_atan.o q_atn1.o    q_atnh.o    q_cos.o q_cos1.o    q_cosh.o    q_cot.o q_coth.o    q_cth1.o    q_ep1.o q_epm1.o    q_errm.o    q_ex10.o q_exp.o     q_exp2.o    q_expm.o q_glbl.o    q_lg10.o    q_log.o q_log1.o    q_log2.o    q_rtrg.o q_sin.o     q_sin1.o    q_sinh.o q_tan.o     q_tanh.o    sin.o sinh.o      sqr.o       sqrt.o tan.o       tanh.o
ar: /home/roberto/filib++/interval/macro/lib/fi_lib.a: No such file or directorymake[2]: *** [/home/roberto/filib++/interval/macro/lib/fi_lib.a] Error 1
make[2]: Leaving directory `/home/roberto/filib++/interval/macro/src'
make[1]: *** [lib] Error 2
make[1]: Leaving directory `/home/roberto/filib++/interval/macro'
Supply variable GMAKE please!
Making primitive.so from primitive.cpp in ieee: done.
Building shared prim support library libprim.so.0: done.
Setting symbolic link from lib/libprim.so.0 to lib/libprim.so: done.

So one has to create the directory .../interval/macro/lib by hand
to complete the compilation of the library.  May I also suggest
to invoke `ar' with the options `cruv' instead of just `rv'?
Note also the confusing error message "Supply variable GMAKE please!".

Another problem is that the distribution seems somewhat mutilated:
in fact, the `test' directory seems missing and this is responsible
for `make check' to fail and for useful things like `make clean'
to print not very reassuring error messages.
That is all for the moment.
Please let me know when you update the distribution (BTW: it would
be very useful to identify archives for different versions like
filib++-2.0.1.tar.gz, filib++-2.0.2.tar.gz and so forth.)
All the best

      Roberto

-- 
Prof. Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara at cs.unipr.it