[LC++]g++ 3.2 and C++ standards

Jack Lloyd lloyd at acm.jhu.edu
Thu Nov 7 05:45:02 UTC 2002

On Wed, 6 Nov 2002, Carlo Wood wrote:

> On Wed, Nov 06, 2002 at 02:12:42PM -0500, Jack Lloyd wrote:
> > Well, maybe. 3.2 has a lot o' bugs, and there are valid ISO C++ programs
> > (mostly ones doing really obscure things) that 3.2 can't deal with.
> > However, in general 3.2 is much more standards-conformant than 2.96 was.
> I disagree, 3.2 has a lot less bugs than 2.96.  And *certainly*
> 3.2 can compile everything 2.96 can - so, if you run into a problem
> with 3.2 (like that the compile crashes) then you most likely are
> dealing with code that 2.96 doesn't compile at all.

Oh, agreed (2.96 was a mess). But your original statement was basically:

Code doesn't compile with 3.2 -> is not standards conforming

which is not strictly true (though I'm sure about 99.9% of the cases where
code was accepted by 2.96 and is not by 3.2 is because the code isn't
valid C++). Actually, there are about 8-10 known regressions of 3.2 as
compared to 2.95.3 or 3.0 (most are getting fixed in 3.2.1 but there are a
few that probably aren't).


More information about the tuxCPProgramming mailing list