Bill Allombert on Thu, 29 Apr 2004 00:40:12 +0200


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: gp: anal.c more bugs


On Wed, Apr 28, 2004 at 12:34:38AM -0700, Ilya Zakharevich wrote:
> On Wed, Apr 28, 2004 at 09:17:13AM +0200, Bill Allombert wrote:
> > > Could you be more explicit: avoid *which* problem?  Is ABI of the (5?)
> > > used X11 functions changing all the time?
> 
> > No, the problem is that you can compile PARI/GP to use a different graphic
> > library (Qt, fltk, none) instead of X11. That should not lead to a libpari
> > incompatible with another.
> 
> Can you explain where you see the difference with the situation with
> GMP kernel?  If there is no difference, I do not see any reason to
> avoid inclusion of the graphing part...

Since GMP-enabled libpari and non-GMP enabled libpari use a different
soname it is possible to install both shared libraries on the machine and
a binary linked with any of them will work since the dynamic linker will
know how to pick the right one. if the library needed is not present, the
dynamic linker will fail right away and not cause random crashes.

I am not comfortable using different soname for every graphic
engine/kernel supported (none,none-gmp,X11,X11-gmp,fltk,fltk-gmp,Qt,Qt-none)
since it would be unwieldly to install all of them.

PARI is a library for performing fast computation in number-theory.
For that purpose, GMP support is a large improvement. On the other end,
plotting support outside of GP is marginaly useful. The documentation
go as far as stating that << plotting is not even a side purpose of PARI >>.

Cheers,
Bill.