Re: USE_LSB_TAG not supported on this platform

[Paul Eggert]

Herbert J. Skuhra wrote:
> INTPTR_MAX=0x7fffffff
> EMACS_INT_MAX=0x7fffffffffffffffLL
>
> Before the commit mentioned above both are 0x7fffffff.

Commit d6585a910ed3e9e0e43c093b5fbfeb6d56b703b4 should not have affected 
EMACS_INT_MAX, and I don't see how it did.

I assume you are not configuring with --with-wide-int, so the reason 
EMACS_INT_MAX has that large value is due to the following tests in src/lisp.h:

  # if INTPTR_MAX <= 0
  #  error "INTPTR_MAX misconfigured"
  # elif INTPTR_MAX <= INT_MAX >> NONPOINTER_BITS && !defined WIDE_EMACS_INT
  ...
  # elif INTPTR_MAX <= LONG_MAX >> NONPOINTER_BITS && !defined WIDE_EMACS_INT
  ...
  # elif INTPTR_MAX <= LLONG_MAX
  ...
  #  define EMACS_INT_MAX LLONG_MAX
  #endif

Is the value of NONPOINTER_BITS 0 on your platform, or 3?  If 0, is this 
correct? That is, does malloc ever return a pointer that, when treated as an 
integer, is not a multiple of 8? If malloc always returns a multiple of 8, then 
we should be able to work around the problem by setting NONPOINTER_BITS to 0 on 
your platform.