[boost] Re: status of 64bit support in integer.hpp?

From: Daryle Walker (dwalker07_at_[hidden])
Date: 2002-12-20 03:35:13

On Wednesday, December 18, 2002, at 9:16 AM, Neal D. Becker wrote:

> What is the status of adding 64bit support to integer.hpp?  I am
> looking for boost::int_t<n>::least for n > 32.

Nothing is happening right now.  I'm sorry if I didn't make this clear 
before: I'm _not_ doing this for you; you have to make the updated 
header and documentation yourself and show it to us for review.  And 
please remember the concerns I brought up on earlier threads.

1.	If 64-bits is already supported by the official C++ integral types, 
then nothing has to be done.

2.	Unfortunately, most compilers (AFAIK) keep int == long == 32 bits 
instead of bumping long up to 64 bits.  Those compilers could make a 
new type for 64 bit integrals.  But remember:

        The existence of these extra types is NOT guaranteed.

Your extension has to fall back safely for a compiler without extended 
integral types.  It won't be accepted otherwise.  (If it was accepted, 
then a conforming compiler that doesn't have extensions wouldn't work.) 
  Maybe the preprocessor could be used here.

3.	The implementations of the types in integer.hpp specialize templates 
for certain numbers.  There are gaps in those numbers; those gaps are 
intentional, to reject too-large bit lengths.  Don't fill in those 
gaps; you have to move them around or completely change the 
implementation.

4.	The CRC library uses the resulting type from integer.hpp as a 
value-based template parameter.  There is no guarantee that extended 
integral types can be used as value-based template parameters (since 
they're unofficial).  Your extension will _break_ CRC.  Since I'm also 
the CRC author, I'm willing to change CRC to be hard-coded to unsigned 
long.  There may be other Boost libraries that make the same 
assumption.  If there are, please let the maintainers of those 
libraries know so they can be part of the discussion.

Daryle