C++ PATCH: Disallow floating-point literals in integral-constant expressions

[Mark Mitchell]

Mike Stump wrote:
> On Jan 31, 2005, at 5:02 PM, Mark Mitchell wrote:
> 
>> The standard says that floating-point literals are only allowed in
>> integral constant expressions if they are immediately cast to an
>> integral or enumeration type.  When I implemented checks for integral
>> constant-expressions, I failed to check this case.  Remedied with the
>> attached patch.
> 
> 
> If users have this in their code, then this can be seen as a removal of 
> a feature by them.  In that case, we should warn about it for a release, 
> then make it an error, or put it under -fpermissive.

We have checks for virtually all invalid integral constant expressions 
in GCC 3.4; we missed this case.  So, we're fixing that bug.

It is not true that if GCC X accepted a piece of code, but GCC X + 1 
does not, then we have removed a feature and must deprecate, etc.  There 
is nothing wrong with diagnosing invalid code, unless the invalid code 
was part of an intentional (and usually documented) GNU extension. 
Other C++ compilers (like EDG) do the same thing: the guarantee from 
release-to-release is that documented extensions and conforming code 
will continue to work, not that all bugs will go through the deprecation 
cycle.

-- 
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com
(916) 791-8304