On 2 April 2011 07:32, Nathan Ridge wrote: > > Doesn't accepting invalid code make a compiler non-conforming? In some cases, but gcc has lots of non-standard extensions and so is intentionally non-conforming in many ways. A valid C++ program that doesn't use those extensions shouldn't be able to tell the difference. That doesn't apply in this case though, I've commented on your bug report pointing out the changes that introduced this behaviour and it was done intentionally to implement the proposed resolution to DR 547, which was voted into the C++0x FDIS at the last meeting.
