Re: [gmp 3.1.1] LONG_MAX and ULONG_MAX redefined

[Torbjorn Granlund]

Vincent Lefevre <address@hidden> writes:

  On Fri, Dec 15, 2000 at 05:31:50 +0100, Torbjorn Granlund wrote:
  > No, no, no.  gmp-impl.h is *not* an application header.
  > It is a GMP-internal header.  We do anything there that is useful
  > for the GMP implementation.
  
  Anyway, not using <limits.h> to get ULONG_MAX, LONG_MAX, etc. (i.e. the
  way specified by the C standard) is *very* bad programming. For instance,
  your code use the fact that bytes have 8 bits:
  
  #define SIGNED_TYPE_MIN(type)        (((type) -1) << (8*sizeof(type)-1))
  
  and is not portable.
  
You and we have different goals.

We want GMP to work on as many systems as possible.  Some old systems
lack limits.h, therefore we don't use it in the pure library (it is
used in some misc files, though).

You get annoyed by purely theoretical problems with the GMP code.  If
we were to spend our time on trying to address such problems, we would
have less time for improving GMP.

Can we please now get on with useful work?  You've made useful
contributions in the past, which we did appreciate.

-- 
Torbjörn