Re: cpp issues

[Jeffrey A Law <law at hurl dot cygnus dot com>]

  In message <199902102008.PAA26023@caip.rutgers.edu>you write:
  >  > From: Dave Brolley <brolley@cygnus.com>
  >  > 
  >  > Kaveh R. Ghazi wrote:
  >  > 
  >  > >          I was in a holding pattern because I couldn't bootstrap with
  >  > > --enable-c-cpplib on Irix6 or OSF4.  I was waiting for either Dave or
  >  > > Zack to fix cpplib.  See:
  >  > >
  >  > > http://www.cygnus.com/ml/egcs-patches/1998-Dec/0421.html
  >  > > http://www.cygnus.com/ml/egcs-patches/1998-Dec/0428.html
  >  > > http://www.cygnus.com/ml/egcs-patches/1999-Jan/0063.html
  >  > 
  >  > As I recall, you were able to bootstrap on other platforms and that your
  >  > patch had nothing to do with the Irix6 or OSF4 failures. I think you sho
  > uld
  >  > submit another patch for us to look at.
  >  > Dave
  > 
  > 
  > 	Fine.  The first step is to agree on how we would define
  > HOST_WIDEST_INT, et al.  Jeff had some comments, e.g.:
  > 
  > 1.  No use of unsigned_HOST_WIDEST_INT, (note the underscore after unsigned
Right.    It should look and work just like HOST_WIDE_INT, with the only
difference that it is the widest integer available on the host platform.

  > 2.  No use of intmax_t.  (Follows from #1 cause we can't say
  > 	"unsigned intmax_t".)
Right.

  > 3.  No checks if int > long.  Ignore these brain damaged non-ansi hosts.
Right.

  > 4.  If long long > long and its available, use it, otherwise use long.
Right.


  > +/* Find the largest host integer type and set its size and type.
  > +   HOST_WIDEST_INT differs from HOST_WIDE_INT in that the WIDE form
  > +   must fit into a register, whereas WIDEST can be larger.  This must
  > +   appear after <limits.h> since we only use `long long' if its bigger
  > +   than a `long' and also if it is supported by macros in limits.h.
  > +   You won't get these defined if you don't include {ht}config.h
  > +   before this file. */
A nit, I wouldn't wire HOST_WIDE_INT to being a form that must fit in a
register.    I'd describe HOST_WIDEST_INT as the widest possible integer
mode supported by the host, which may (or may not) be wider than
HOST_WIDE_INT.


The patch looks pretty good otherwise.  Update the comment and please
install this patch.

jeff