from Dave Thomas on 2001-08-05 at 02:23:16:
> Eli Green <eli.green / codedogs.ca> writes:
> 
> > I don't think rubicon is 100% there yet ... with the latest rubyunits and
> > rubicon, even my FreeBSD machine gives off some weird errors while doing a
> > make test -- and ruby runs fantastically under FreeBSD.
> 
> Rubicon definitely isn't there under any OS other than Linux, for the
> basic reason that I haven't done any work in getting in consistent on
> those platforms. I'm just starting to experiment with getting it to
> run under W2K/Cygwin, and in the process found some hidden incorrect
> assumptions in some of the tests. These I've fixed. I'm sure that the
> BSD series will reveal more of these, and I'd be very, very happy to
> integrate patches to rectify these. All I ask is that we don't
> eliminate a test simply because it doesn't run: let's first establish
> whether it _should_ run.

Woops. Sorry if I came off sounding like I didn't think rubicon was any good
... I'm not about to slander software with a 0.1 after its name. =)

In fact, what I'm basically doing is running it against FreeBSD, and making
sure that the QNX port is at least as error-free as the FreeBSD port. As I
said, most of the problems I've encountered are under TestIO and stuff that
uses threads or signals, which is always finicky.

So despite its infancy, it's still being quite useful.

It's bloody disconcerting to get assertions on TestFixnum, though. =)

> I'm very keen to build rubicon into an effective verification suite
> for Ruby across all supported platforms. I feel that it will help
> greatly to have some kind of centralized place we can reference to
> answer questions about what runs and what doesn't run on a particular
> OS. If anyone wants to contribute to this, please feel free. The
> latest rubicon is in the src/ CVS repository.

K, a couple of quick comments:

"telldir" throws NotImplemented on platforms that don't have the call, which
results in 2 errors. Technically, these really shouldn't be errors, since
there's no fix for it. =)
(what's really aggravating is that QNX has telldir in its include files, just
 not in libc ... weirdos...)

I get those "frozen object" messages a fair bit under QNX, and a lot under
FreeBSD. The actual error is a TypeError ... anybody know the exact source of
this exception?

> Many thanks

Many thanks to you for writing it in the first place.

> Dave