On Mar 22, 2012, at 6:15 AM, Michael Matz wrote:
> That certainly is strictly better than any of the other possibilities, I
> just didn't get the impression from your second mail to this thread that
> you were even considering doing that. Good I was wrong.
All I wanted, was to remove the assert... The review point was, no, not unless
you fix the issues so we don't get wrong code-gen and could you make it sign
extending as well? So, sure, sounds reasonable to me. I was going to do the
work in the end, just didn't plan on doing it today. Today, tomorrow, not
worth quibbling over the exact date the work is done. But, my final point is,
the assert is wrong, and it has to go, and (almost) gone it is. :-) I'm happy.
> I would call it too strict, not wrong.
Do you have users? Have you ever told them the compiler isn't wrong when it
ICEs for perfectly valid code? I've never done that before, and never plan to,
no one has convinced me it is the right approach. If you want me to not use
the term wrong, you'd need to furnish a web site that somehow proves your
point. Wrong is what I use when that the compiler does is wrong. It is that
simple. Failing to compile valid code, is wrong.
> Because there are (or were after
> your fixes get it) values where there was a problem. Of course that's
> again just splitting hair about terminology :)
Yeah, I'm not into hair splitting on terminology. I'm more into actual
functionality of the compiler to the end user.