preprocessor/5806: The preprocessor evaluates expression s in 64-bit, violating IS C++ 16.1.4

[Dana, Eric]

The following reply was made to PR preprocessor/5806; it has been noted by GNATS.

From: "Dana, Eric" <Eric_Dana@bmc.com>
To: "'Zack Weinberg'" <zack@codesourcery.com>,
   "'Neil Booth'"
	 <neil@daikokuya.demon.co.uk>
Cc: "Dana, Eric" <Eric_Dana@bmc.com>,
   "'gcc-gnats@gcc.gnu.org'"
	 <gcc-gnats@gcc.gnu.org>,
   "Schoeller, Dick" <Dick_Schoeller@bmc.com>
Subject: RE: preprocessor/5806: The preprocessor evaluates expression s in
	 64-bit, violating IS C++ 16.1.4
Date: Thu, 14 Mar 2002 14:58:41 -0600

 Zack,
 
    Is there anything that can be done for 3.1? In it's present state,
    Dynix cannot work properly as many of its include files test for
    32/64 bit using the preprocessor. For example:
 
 	#if ((~0UL) == 0xffffffffUL)
 	typedef unsigned int    size_t;         /* ILP32 size_t */
 	#else /* ((~0UL) == 0xffffffffUL) */
 	typedef unsigned long   size_t;         /* LP64 size_t */
 	#endif /* ((~0UL) == 0xffffffffUL) */
 
    The code between the #if #else #endif varies depending on the header
 file.
    For 3.1, could a modification in fixinc be made to handle the above
    mentioned type of case? Fixinc would know what the size of the HOST
    long is (can it determine the size of the TARGET long?).
 
 --Eric--
 
 -----Original Message-----
 From: Zack Weinberg [mailto:zack@codesourcery.com] 
 Sent: Thursday, March 14, 2002 2:39 PM
 To: Neil Booth
 Cc: eric_dana@bmc.com; gcc-gnats@gcc.gnu.org; dick_schoeller@bmc.com
 Subject: Re: preprocessor/5806: The preprocessor evaluates expression s in
 64-bit, violating IS C++ 16.1.4
 
 
 On Fri, Mar 01, 2002 at 11:13:14PM +0000, Neil Booth wrote:
 > Zack Weinberg wrote:-
 > 
 > > The preprocessor ought to do this only in C99 mode; in C++ and
 > > presumably also C89 mode (I don't have a copy of C89 to check) it
 > > should use long/unsigned long, which may be 32 bits wide on some
 > > targets.  Unfortunately, doing this properly requires substantial
 > > changes to GCC, which are planned, but not currently practical.
 > > (We're already not quite compliant; preprocessor arithmetic uses the
 > > _host's_ idea of intmax_t, not the target's.)
 > > 
 > > Neil - as a stopgap, it occurs to me that we could mask intermediate
 > > values down to 32 bits when in C89/C++ mode and sizeof(target unsigned
 > > long) == 4.  Thoughts?
 > 
 > Sounds OK to me.  Like you, I have the C++ and C99 standards, but no C89
 > lying around.  I do believe that the C++ preprocessor section was copied
 > almost verbatim (with some changes for bool and true / false) from C89
 > though.
 
 I looked into this a little, and it founders on the usual problem:
 LONG_TYPE_SIZE can vary with target variables.  So this is blocked on
 the integrated -E mode.  We should make a serious effort to get that
 done for 3.2.
 
 zw