[C/C++] PR 13358 long long and C++ do not mix well

[Mark Mitchell]

Manuel López-Ibáñez wrote:

> If one changes a message and there are not testsuite changes, then one
> cannot know whether this message is actually ever tested.

That's clearly true, at some level, but I just think it's overly
pedantic.  It's not like messages change by themselves. :-)  In other
words, I think there's something qualitatively different between a
language feature (where a change to some other part of the compiler
might accidentally break it) and a change to an error message, which is
generally being done very consciously.

I just feel that the benefits of avoiding the potential error message
breakage are outweighed by the costs.  Since it's a tradeoff, reasonable
people can certainly disagree.  I'd like to see us aim for a middle
ground where we have enough text to capture key concepts, but not the
entire message.

Thanks,

-- 
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713