Re: bignum branch

[Andy Moreton]

On Mon 06 Aug 2018, Tom Tromey wrote:
> As far as I know the branch is ready.  I have two uncommitted patches:
> one for dumping (just sent) and one to use mpz_import (that I can't
> test).  The second one seems like a minor improvement and could probably
> be dropped.

Have you run "make check" on the bignum branch ?

I see various test failures (many of which are also failures on master),
and three crashes:

1) A crash in the json tests
   This is bug#32381, fixed on master in commit 3eac378c96.

2) An eassert in make_bignum_str called from string_to_number.
   This appears to be caused by differences in the syntax accepted by
   emacs lisp and by mpz_set_str.  To reproduce:

     ELISP> (format "+%S" (- most-negative-fixnum))
     "+2305843009213693952"
     ELISP> (string-to-number "+2305843009213693952")
     ### crash ###

   It also appears that make_bignum_str in alloc.c is only used in 
   string_to_number, and so should be open coded in its caller.

3) Another crash which I have not yet had time to analyse with gdb.

    AndyM