Bug#724320: gmp: x32: sizeof(mp_limb_t)!=sizeof(void*) is not supported by GAP and PARI

[Bill Allombert]

On Mon, Sep 23, 2013 at 12:55:18PM -0500, Steve M. Robbins wrote:
> Hello Bill,
> 
> I'd like to understand what you believe is the bug.

Note that I set the severity to wishlist.

> On Mon, Sep 23, 2013 at 07:02:41PM +0200, Bill Allombert wrote:
> 
> > On x32, GMP set sizeof(mp_limb_t) to 8, see
> > <http://buildd.debian-ports.org/status/fetch.php?pkg=gmp&arch=x32&ver=2%3A5.1.2%2Bdfsg-3&stamp=1379197327>
> 
> Is 8 the wrong value?  If so, any idead why the GMP unit tests don't fail?

GMP can be configured to use either 4 or 8. Both are correct GMP-wise.
(The same is true on standard i386, except using 8 requires the processor
to support the x86_64 extension, which is already assumed on x32).

> > However this is incompatible with software that requires sizeof(mp_limb_t)==sizeof(void*),
> > like pari and gap, which then FTBFS on x32.
> 
> Is this a bug in pari and gap or in gmp?

Admittedly, this is a bug in pari and gap for having such requirement. However,
it seems unlikely to be ever fixed, hence this report.

Using 8 should result in a faster library, so this is a trade-off, even if I
do not expect that people will use x32 for HPC.

Cheers,
Bill.