Bill Allombert on Wed, 28 Jan 2004 02:10:48 +0100


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

Preparing PARI 2.3


Hello PARI-dev,

Since we are speaking about `release', I would like to see what need
to be done.

Above is a mail I sent a year ago that listed some problems that needed
to be addressed before releasing.

On Thu, Dec 12, 2002 at 08:02:31PM +0100, Bill Allombert wrote:
> There are several problesm to be addressed in the development release:
> 
> --- Fix the problem with functions returning int on 64bit architecture.

Done recently.

> --- The polrootspadic(x^2+8*x+4, 2, 2) bug is still not fixed.

It is rather annoying, since PARI 2.1 give the correct result while 2.2
give another (incorrect) result.

> --- Fix problems with C++ compilers. (almost done).
Done, as can testify http://pari.math.u-bordeaux.fr/buildlog.html
(thanks to all involved, by the way).

> --- Fix the ABI/API: define public functions in paridecl.h, not just those
> we need to build PARI.

That need to happen, though latest changes have improved a bit on the
situation.

I plan to review the "regular" interface (FpX, Flx, vecsmall,etc.) to
see what functions need to be added to paridecl.h

> --- GP2C support: ship the func.desc GP2C file with PARI so that gp2c 
> can be compiled without access to PARI source.

This is done.

> --- add provision to add MacOS X support in a minor revision.

I believe it is done. Do the MacOS X users agree ?

> --- a lot more I cannot remember now :)

This one allow me to complete the list:

--- Change the soname of libpari.so.2 to libpari-gmp.so.2 when build
with GMP support (as discussed previously).

--- Fix some minor impler in the GMP kernel : 
    bittest(a,b,-c) is not tested (and wrong).

--- Provide inline equivalent of macro so that gp2c work better

--- Add more description of interface for gp2c (in particular to
reflect the ulong->pari_sp and lgef->lg change.)

--- Add substpol() and reinstate old subst behaviour.

--- Get rid of the gnuplot support due to licensing issue.

--- and of course others I will remember later :)

But it seems that overall we have made a large progress toward a releasable
version during the last year, in particular with the rewrite by Karim of
the Configure system.

Cheers,
Bill.