- GiNaC-list - ginac.de

cln 1.1 does not with with -fno-exceptions?
by Roland Richter 12 Jan '01

12 Jan '01

Dear all, I recently tried to install cln 1.1 under SuSE Linux 7.0 When adding the -fno-exceptions flag to CXXFLAGS, as recommended in the INSTALL file, everything compiled fine, but then the test (cln-1.1/tests/exam) results in a seg fault. Making everything without that flag worked fine. - Roland - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

2 1

Release 0.7.0 is on the wire
by Richard B. Kreckel 15 Dec '00

15 Dec '00

Hi all, I have just finished packaging GiNaC 0.7.0. It is sitting at the usual places, see <http://www.ginac.de/Download.html>. Beware that you really do need CLN 1.1 (silently released yesterday) in order to use it! This is so, because CLN 1.1 is interface-incompatible (namespace cln) with earlier versions of CLN and GiNaC 0.7.0 is expecting the new interface. For those of you (Eduardo, Bernard) who want to extract a cln::cl_N from GiNaC::numeric for various reasons, there is now a method GiNaC::numeric::to_cl_N() which does right that. (I was first having a casting operator in mind, but that turned out to lead into ambiguities quite easily so we'll use .to_cl_N() from now on.) Cint users should upgrade to at least 5.14.62 in order to use GiNaC-cint with the current release. So, this is the list of required package versions and the order of steps how to proceed: * Install GMP >= 3.0 (current is 3.1.1) (optional) * Install CLN >= 1.1 (current is 1.1) * Install Cint >= 5.14.62 (current is 5.14.65) (optional) * Install GiNaC 0.7.0 * rejoice, cheer, sing... ;-) We plan to provide .deb and .rpm packages next week. And please read the file NEWS to see what else has changed. Happy hacking -rbk. -- Richard Kreckel <Richard.Kreckel(a)Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/> - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

1 0

Extracting only the integral (or fractional) part of a number
by Eduardo M Kalinowski 08 Dec '00

08 Dec '00

Hi, I have another question: how can I extract only the integral (or fractional) part of a number? ie, how to get "3" and "0.141593" from "3.141593"? -- Common sense is the collection of prejudices acquired by age eighteen. -- Albert Einstein Eduardo M Kalinowski (ekalin(a)iname.com) http://move.to/hpkb http://cami-ufpr.hpg.com.br - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

1 0

Extracting a numeric from an ex
by Eduardo M Kalinowski 08 Dec '00

08 Dec '00

Continuing the previous message... I would find it extremely useful if it was possible to get a numeric representing the ex's contents, _if_ that is possible, of course. Specifically, it would a be a function like this: class ex { .... numeric to_numeric() const; .... } Or any other significant name. Of course, there are times when a numeric cannot represent an ex. In these cases, the most appropriate solution seems to throw an exception, but other ways of allerting the user are possible. Another function, returning a bool saying wheter the ex can be converted into a numeric would also be nice. Is the above possible? -- The church is near but the road is icy; the bar is far away but I will walk carefully. -- Russian Proverb Eduardo M Kalinowski (ekalin(a)iname.com) http://move.to/hpkb http://cami-ufpr.hpg.com.br - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

3 3

Digits variable ignored when parsing strings?
by Eduardo M Kalinowski 07 Dec '00

07 Dec '00

Hi, It seems the value of the 'Digits' variable is ignored when a GiNaC::numeric is created from a string. This behaviour can be observed in ginsh, for example. The created number always has something like 17 digits. I suppose this isn't what should happen. -- God created woman. And boredom did indeed cease from that moment -- but many other things ceased as well. Woman was God's second mistake. -- Nietzsche Eduardo M Kalinowski (ekalin(a)iname.com) http://move.to/hpkb http://cami-ufpr.hpg.com.br - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

2 2

Equivalent of evalf for numeric (for fractions)
by Eduardo M Kalinowski 06 Dec '00

06 Dec '00

Hi again, How can I convert a GiNaC::numeric that contains a fraction (or even a integer, for that matter) into a numeric with the real representation of the fraction? Currently I'm adding "0.0" to the numeric to force it into a real number. This works, but is ugly. How about an "evalf" function for GiNaC::numeric? -- I'm always looking for a new idea that will be more productive than its cost. -- David Rockefeller Eduardo M Kalinowski (ekalin(a)iname.com) http://move.to/hpkb http://cami-ufpr.hpg.com.br - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

2 2

floor function for GiNaC::numeric
by Eduardo M Kalinowski 11 Nov '00

11 Nov '00

Hi, I've looked through the documentation and header files for GiNaC::numeric and I haven't been able to find the floor (or ceil, for that matter) function for that class. Have I overlooked something, or is that missing? How can I implement one? I know CLN has a floor1 function that is what I want. -- So far we've managed to avoid turning Perl into APL. :-) -- Larry Wall in <199702251904.LAA28261(a)wall.org> Eduardo M Kalinowski (ekalin(a)iname.com) http://move.to/hpkb http://cami-ufpr.hpg.com.br - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

2 2

collect problem
by Ben Sapp 20 Oct '00

20 Oct '00

Hi, I have program that crashes when it uses the collect function of the ex class. The strange error I receive is: Not a 32-bit integer: -9223372028264841216 I am able to print both the expression and the symbol I am going to use for the collect call before hand and all seems fine. I archived the symbol and expression just before I knew it would crash and I can look at the function with viewgar. However if I make a program that uses collect on the expression I get the same strange error and the program exits. Is this a bug in GiNaC? Is it a bug in my software? What could be causing this? How can I fix it? I have attached the archive file and the source to the program that reads the archive file. I did not give the source to the software that originally produced it because it is a function that is dynamically linked with another program.(GNU Octave) In case my system configuration is important: [ben@pcben Linux]$ ginac-config --version 0.6.4 [ben@pcben Linux]$ g++ -v Reading specs from /usr/local/lib/gcc-lib/i686-pc-linux-gnu/2.95.2/specs gcc version 2.95.2 19991024 (release) [ben@pcben Linux]$ uname -a Linux pcben 2.2.12-20 #1 Mon Sep 27 10:40:35 EDT 1999 i686 unknown Thanks for any insight. -- Ben Sapp Los Alamos National Laboratory email: <mailto:bsapp@lanl.gov> Phone: (505)667-3277 Fax: (505)665-7920 URL: http://www.neutrino.lanl.gov/ -- GARCclasssymbolnamexThe symbolmuladdrestnumericnumber1coeffR-1 11529215046068469760 -61overall_coeffpowerbasisR1 9223372036854775808 -62exponentR1 13835058055282163712 -62The expression J J;[kJ{<J;[k;[; [k #include <ginac/ginac.h> #include <iostream> #include <fstream> using namespace GiNaC; int main() { archive a; ifstream in("foobar.gar"); in >> a; symbol x("x"); lst syms(x); ex expr = a.unarchive_ex(syms,"The expression"); cout << "expr = " << expr << endl; cout << expr.collect(x); return(0); }

4 4

Re: Ben Sapp's Problem
by Richard B. Kreckel 20 Oct '00

20 Oct '00

Hi, Sorry, Christian, for putting this into the list so late but I thought Ben and others might wish to know what is going on. And sorry to all the others for the German in it... On Fri, 20 Oct 2000, Christian Bauer wrote: > Ich habe jetzt eine Version gebastelt (noch nicht im CVS), die > nicht-integer-Exponenten bei (l)degree() und coeff() als 0 ansieht, d.h: What I was thinking of, was that ldegree returns the corresponding floor and degree the corresponding ceiling. Is this idea somehow screwed or does it just not matter at all? > > e=5*x^2.0+42*x^3; > 5*x^(2.0)+42*x^3 > > degree(e,x); > 3 > > ldegree(e,x); > 0 > > coeff(e,x,0); > 5*x^(2.0) > > coeff(e,x,2); > 0 > > coeff(e,x,3); > 42 > > collect(e,x); > 5*x^(2.0)+42*x^3 [...] > Etwas inkonsistent finde ich, das in GiNaC > > > is(2.0==2); > 1 > > wo doch 2.0 eigentlich ein inexakter Wert ist. > > Wohingegen > > > x^2.0*x^3; > x^(5.0) > > in Ordnung ist. Hmm, there is no clever way in CLN to do just that, one would have to resort to some (trivial) RTTI-fiddling here. Is this just inconsistent or does this lead to potential problems somewhere? -richy. -- Richard Kreckel <Richard.Kreckel(a)Uni-Mainz.DE> <http://wwwthep.physik.uni-mainz.de/~kreckel/> - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

1 0

still problems
by Tilman Utz 19 Sep '00

19 Sep '00

Hi folks, once again I have to ask for your help and advice, although I some moments of success, the great breakthrough is still missing. A lot of my problems in making the library (cln and ginac) resulted from the Windows habit of changing the case of file- and directorynames, e.g. from gmp.h to Gmp.h, so the Makefile didn't find the file. How about you, do you have problems with that, too, and how to prevent it? Well, I changed the filenames by hand and might have missed some. So, the problems I encounter with configure may be a result of that. Here's, what I did and what I got: No problems with installing cygwin 1.1.4 cln-1.0.3 was unzipped and untared to /usr/local, path changed to /usr/local/cln-1.0.3 ./configure --prefix=/usr/local after a while, configure aborts with the message, that it couldn't determine host-type, because config.guess was missing (./config failed for gmp), well, the file is there, but not in ./gmp but in ./autoconf. I move config.guess and later config.sub to ./gmp, then, configure finished without error messages, but in the config.log-File, there are still messages, which are like below: configure:1894: alloca.h: No such file or directory configure: failed program was: #line 1893 "configure" #include "confdefs.h" #include <alloca.h> int main() { char *p = (char *) alloca(2 * sizeof(int)); ; return 0; } similar for filenames like that: fpu_control.h (I can send you the config.log if it might help...) trying the ./configure command again, these errormessages disappear... After executing about half an hour of the Makefile, make stops editing a bunch of errors like: ../src/.libs/libcln.a(cl_F_extendsqrtx.o)(.text+0x611):cl_F_extendsqrtx.cc: unde fined reference to `_GLOBAL_$D$cl_module__cl_ieee__firstglobalfun' ../src/.libs/libcln.a(cl_LF_scale_I.o)(.text+0x532):cl_LF_scale_I.cc: undefined reference to `_GLOBAL_$I$cl_module__cl_ieee__firstglobalfun' ../src/.libs/libcln.a(cl_LF_scale_I.o)(.text+0x539):cl_LF_scale_I.cc: undefined reference to `_GLOBAL_$D$cl_module__cl_ieee__firstglobalfun' ../src/.libs/libcln.a(cl_F_floor22.o)(.text+0x4a6):cl_F_floor22.cc: undefined re ference to `_GLOBAL_$I$cl_module__cl_ieee__firstglobalfun' These occur while the Linker (ld is a linker, isn't it?) is working Within Linux I'm slightly better off. Although I get the config.log errors like above, too, making, checking and installing cln and Ginac works without obvious errors. The GiNaC tests are passed, too, compiling the example program from the GiNaC documentation, works too, but executing the file results in the error that lbginac.0.6.so.3 is not found (although it's in /usr/local/lib). Do you have an idea, what's wrong with my system or the way I try to build the libraries? I have a PentiumIII 600 128MB of RAM, Windows98 and cygwin 1.1.4 installed. Is it possible to transfer the compiled libraries from Linux to cygwin (I mean, gcc is gcc, isn't it?) because I'm more hopeful to bring the libraries to work with LINUX than with Windows... Thanks a lot in advance Tilman Utz - To UNSUBSCRIBE, email to ginac-list(a)ginac.de with a subject of "unsubscribe".

2 1