RE: [boost] VC++ 7.0 Challenge

From: jsiek_at_[hidden]
Date: 2000-08-22 13:06:14

Note that I did not complain about their non-conformance per se, but
about their "refusal to conform". I have no problem with compilers
that are non-conformant (all of them) but that are working towards
conformance (EDG, Metrowerks, g++, etc.).

Bill Wade writes:
 > > -----Original Message-----
 > > From: jsiek_at_[hidden] [mailto:jsiek_at_[hidden]]
 > >
 > > Is there anything the ISO can do about refusal to conform?
 > 
 > Do you know of any conforming compilers?  I haven't found any where 'export'
 > works.
 > 
 > The last time I tried
 >    char* p = new char[size_t(-1L)];
 > on a 32-bit linux platform it returned a non-null pointer, but the pointer
 > did not point at 2^32-1 usable characters.
 > 
 > I think the state of the art for determining C++ conformance is to get a
 > statement by a trusted source about how well feature X works on compiler Y.
 > In many cases "a trusted source" may have to be a test program that you
 > write yourself.

Very true... lots of us do this. What we really need is a common data
base of code snippets that we can all contribute to. A number of C++
libraries have collections of these code snippets that get run during
configure (STLport, Blitz, MTL, GGCL, etc.), though there is no
process in place for sharing this important resource.

Cheers,

Jeremy