do we expect compiler to catch this?

[Brooks Moses]

Tobias.Schlueter@Physik.Uni-Muenchen.DE wrote:
> If the code you gave is also in a single file, the compiler could (and IMO
> should) catch this, that's why I referred to quality of implementation.

It occurs to me that, in the general case, there will be quite a number 
of tests that fall under the category of "the compiler is not required 
to catch this, and it currently doesn't catch it, but it would be nice 
if it did."

Thus, I'd propose having two somewhat separate testsuites:

(1) A "Constraints" testsuite of things the compiler is required to 
catch, either because the standard requires it or because it already 
catches it and a failure to do so is a regression.

(2) A "Quality of Implementation" testsuite of things that it would be 
nice if the compiler caught, but it's not required to catch.

One would expect that the first testsuite should not show failures, 
while that the second testsuite will normally not show successes (if it 
does, they get moved to the first testsuite).

Obviously, the second testsuite would not be used in regression testing, 
but I think it would still be quite useful to have around as a "wish 
list" of sorts.

- Brooks