Re: math::pari + pari-2.1.2

[Karim BELABAS <Karim.Belabas@math.u-psud.fr>]

On Thu, 25 Oct 2001, Ilya Zakharevich wrote:
> > 2) I had many configuration problems, all boiling down to Math::PARI's using
> > perl's Config.pm when our perl installation was installed from a binary
> > package [frequent situation on Solaris]
[...]
> Junk in, junk out.  This is a Perl module; if the Perl configuration
> is busted bad enough, one is not able to build a Perl module, period.
> Nothing specific to Math::Pari.  I do not think I *want* to treat this
> case specially.

Sure. A quick mention of the symptoms in the README would be user-friendly.

> Some time I will ask you to run some tests, since I *am* testing the
> headers...

OK.

> BTW, any reason why ulong is *still* used in the sources?  I sent a
> recipe how to fix it around 5 years ago.  IIRC, it was something as
> simple as a global edit as in
>
>   pfind src '=~ s/\bulong\b/PARI_ulong/'

Well, it's a (nearly) standard type name. And it's easy enough to check
whether it already exists or not. It only becomes a problem when you include
whole new sets of headers in order to link with other libraries. But then
PARI namespace is hardly protected...

Also it's not that simple, it would make a lot of sense to have avma and
friends be of type e.g PARI_stack_ptr, whereas the arithmetic word operations
should operate on e.g PARI_word [these two changes would remove zillions of
Warnings from picky compilers, btw] etc... It doesn't make much sense to make
a global edit which would not clarify things, falling back to (unsigned long)
would do just as well.

>> glob failed (child exited with status 1) at Makefile.PL line 97.
>
> Your perl is not correctly compiled.  Did you run `make test'?

Well, that part of my mail refered to the (badly) pre-installed one.
Everything's fine with my private perl install.

> > Pari.xs:224: warning: static declaration for `changevalue' follows non-static
>
> Thanks.  Cannot find it in CHANGES, when was it made global?  It is
> not in 2.0.15 yet...

anal.c revision 1.55 (2001/09/18 11:04:50) that means just after
release-2.2.0.  It was a _huge_ patch [ introduced writebin() ], I forgot to
mention the globalization in CHANGES (also in 'cvs log'...).

> > 4) make test fails miserably:
>
> This is typical for a configuration change without makeing realclean
> (I'm adding this to README).  Underscore problem.  Please re-report.

OK, after 'make realclean', everything works like a charm.

Thanks!

    Karim.
-- 
Karim Belabas                    email: Karim.Belabas@math.u-psud.fr
Dep. de Mathematiques, Bat. 425
Universite Paris-Sud             Tel: (00 33) 1 69 15 57 48
F-91405 Orsay (France)           Fax: (00 33) 1 69 15 60 19
--
PARI/GP Home Page: http://www.parigp-home.de/