Migration: fplll: 4 to 5

[jgmbenoit]

Attempt to migrate from fplll 4 to fplll 5. Hacked on Debian Sid/experimental box.
Note that the APPLE part was commented out.
Otherwise, some hacks can certainly be improved, in particular the one in
void SET_INTOBJ(Z_NR &v, Obj z)

[laurentbartholdi]

Hi!
I rewrote it (merging your changes), because your PR is not compatible with fplll version 4.

Hopefully the new branch below works for both. Can you check it?

https://github.com/gap-packages/float/tree/jgmbenoit-debian/0.7.4+ds-2/migration/fplll/4to5

[jgmbenoit]

Hello, my PR was purposely oriented towards version fplll 5.
I have just try the rewritten material on a Sid box (fplll 4) and on a Sid/experimental box (fplll 5).
Without surprises, the autogen.sh works well, and the following configure works as well:
$ ./configure --with-gaproot=/usr/lib/gap --with-mpfr=yes --with-mpc=yes --with-mpfi=yes --with-fplll=yes --with-cxsc=no

The build (make) with fplll 4 works, but not the one with fplll 5: I get the error I got yesterday:
fplll.C: In function 'void SET_INTOBJ(fplll::Z_NR&, Bag) [with Z = __mpz_struct [1]; Bag = long unsigned int*_; UInt = long unsigned int]':
fplll.C:49:16: error: invalid conversion from 'mpz_ptr {aka _mpz_struct}' to 'long int' [-fpermissive]
v = mpz_MPZ(MPZ_LONGINT(z));
~~~~~~~^~~~~~~~~~~~~~~~
In file included from /usr/include/fplll/nr/nr.h:15:0,
from /usr/include/fplll/nr/numvect.h:19,
from /usr/include/fplll/nr/matrix.h:21,
from /usr/include/fplll/util.h:23,
from /usr/include/fplll/fplll.h:25,
from /usr/include/fplll.h:1,
from fplll.C:29:
/usr/include/fplll/nr/nr_Z_mpz.inl:89:13: note: initializing argument 1 of 'void fplll::Z_NR::operator=(long int) [with Z = __mpz_struct [1]]'
inline void Z_NR<mpz_t>::operator=(long a) {

Hence my tricky code. I think it is a fplll issue in the sense that it cannot play with some tricky gmp code, but I am not familiar with C++ to be 100% certain.

[jgmbenoit]

Please let me know whether or not you can reproduce the issue ?
Personally, I think it is rather a fplll API issue, that is, a bug report might be emitted. What do you think ?

[laurentbartholdi]

Yes, I can see the same problem. I didn't try different compilers, though.
In all cases, the current float master uses your workaround.

On Wed, 19 Oct 2016 at 03:08 Jérôme Benoit notifications@github.com wrote:

Please let me know whether or not you can reproduce the issue ?
Personally, I think it is rather a fplll API issue, that is, a bug report
might be emitted. What do you think ?

—
You are receiving this because you commented.

Reply to this email directly, view it on GitHub
#13 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/ACIIUF0IA3RsCpDXM1ptE_nyPAfP0hCqks5q1W13gaJpZM4KXmpy
.

Laurent Bartholdi
DMA, École Normale Supérieure, 45 rue d'Ulm, 75005 Paris. +33 14432 2060
Mathematisches Institut, Universität Göttingen, Bunsenstrasse 3-5, D-37073
Göttingen. +49 551 39 7826

[jgmbenoit]

Ok. So, I emitted an issue to fplll