[m-dev.] Test cases (was: for review: big ints)

[Fergus Henderson]

On 08-Apr-1998, Tyson Dowd <trd at stimpy.cs.mu.oz.au> wrote:
> Some regression tests are really just missing "feature tests".  Some
> regression tests are combinations of code that haven't been tested
> together (so never really fit with one "feature" or another -- usually
> they get lumped in with the feature that was changed to fix the bug). 
> But many (most?) regression tests are caused by failure to keep invariants
> invariant, which is not so much a feature as "structural integrity".
> 
> It's the last category that deserve a category of their own, if only
> so future failures don't falsely implicate the wrong part of the
> compiler.
...
> Sometimes.  Some of them, I'll agree with you.  But if a bug in
> maps means that a table generated in polymorphism causes a software
> error in code_gen when compiling some code that uses typeclasses,
> where should the regression test go?

OK, so long as the README makes it clear that not all
regression tests go in the `regressions' subdirectory,
it would be OK to have a `regressions' subdirectory for
regression tests that don't fit cleanly elsewhere.

> Do you think it is worthwhile doing this? Since the way you've
> structured it makes it mainly a quick directory move, it would probably
> be relatively easy.

Yes, I think it would be worthwhile.
However, I'm not volunteering! ;-)
There are of course a lot of other worthwhile tasks around...

-- 
Fergus Henderson <fjh at cs.mu.oz.au>  |  "I have always known that the pursuit
WWW: <http://www.cs.mu.oz.au/~fjh>  |  of excellence is a lethal habit"
PGP: finger fjh at 128.250.37.3        |     -- the last words of T. S. Garp.