Re: GMP memory allocation and Macaulay2

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

* Doug Torrance [2016-05-19 20:58]:
> Currently, PARI sets the GMP memory allocation functions in
> pari_kernel_init().  However, this may cause problems for applications which
> use both the PARI and GMP libraries such as Macaulay2.  Building Macaulay2
> against unpatched GMP and PARI libraries causes segmentation faults.
> 
> Currently, Macaulay2 builds its own GMP (actually MPIR, a drop-in
> replacement) in which mp_set_memory_functions() does nothing.  This is not
> ideal, however.  For example, this is the main problem keeping Macaulay2 out
> of Debian, as it cannot be built with the current Debian GMP and PARI
> packages.  (See [1] for some discussion on this matter.)
> 
> Would it be possible to remove the calls to mp_set_memory_functions() from
> PARI?

We can provide an additional flag in pari_init_opts()'s init_opts
bitmask then use something like

  if (!(init_opts & INIT_noKERNEL_INITm)) pari_kernel_init();

in there. And similarly in pari_close_opts.

You'd juste have to replace your pari_init() call by

  pari_init_opts(..,.., ..|INIT_noKERNEL_INITm)

(or add the INIT_noKERNEL_INITm bit to your existing pari_init_opts call).

I don't foresee unwanted ѕide effects *provided* you
- either leave SIGINT alone and let it be fatal to the whole program,
- or also replace GMP's memory allocation by a custom memory allocator
  that temporarily blocks BIGINT. E.g. gp sets SIGINT to stop whatever
  we were doing to longjmp() to gp's main loop (and present a prompt
  to the user); if we do that within malloc/realloc/free, we can corrupt
  malloc's internal state (=> crash later). You want to avoid an
  analogous issue.

Cheers,

    K.B.
--
Karim Belabas, IMB (UMR 5251)  Tel: (+33) (0)5 40 00 26 17
Universite de Bordeaux         Fax: (+33) (0)5 40 00 69 50
351, cours de la Liberation    http://www.math.u-bordeaux.fr/~kbelabas/
F-33405 Talence (France)       http://pari.math.u-bordeaux.fr/  [PARI/GP]
`