unit test macros

[jesse99]

Using asserts within unit tests is entirely unsatisfactory: when they fail you can't see why without editing the code or breaking out a debugger, and it's difficult to be intelligent about pointing out where the problem is (e.g. maybe the second element of a vector doesn't match the expected value).

Seems like it would be possible to use the new whiz bang macro system to address this. Lots of unit test frameworks out there, but it's probably best to look at one for a statically typed language like Nunit.

[jesse99]

Blocked by #3066 (line! reports line numbers within the macro).

[catamorphism]

#3066 is still open, so this is still blocked.

[graydon]

we have assert_eq and such now. I think the minor issues in #3066 aren't blocking this, and in any case appear to have been fixed by #4888 . Can we close this?

[pnkfelix]

I agree with Graydon that #3066 appears to be fixed. I'm going to double check that our existing unit test patterns produces useful error messages.

[pnkfelix]

Okay, we do indeed get useful stack traces from assertion failures, in particular assert_eq prints out the two values separately.

I don't think we need to adopt a more full-featured test infrastructure without some different motivating examples. (we might consider refining the assert_eq output to narrow in on the relevant bit of substructure that does not match, but that is not this ticket.)