Re: [Gcl-devel] Updates to test suite

[Paul F. Dietz]

Camm Maguire wrote:

> Greetings!  Paul, I've got a patch just about ready to commit
> addressing many of these.  I'm uncertain about one aspect.  GCL
> currently compiles functions to check the number of actually supplied
> arguments at runtime.  Basically, if &allow-other-keys or &rest
> explicitly appears in the lambda-list of the defun, the upper-limit
> check on the supplied arguments to the function at runtime is bypassed.
> My understanding is that this bypass should always be in place, though
> it seems a shame not to be able to declare a function in such a way as
> to have an upper-limit on the number of supplied arguments.

If there's a &rest or &key in the ordinary lambda list then
there is no limit to the number of parameters that may be
safely passed.  Remember, a keyword argument may appear more
than once, and the &rest list can be arbitrarily long.

> The only other residual allow-other-keys errors involve keys that are
> really symbols, i.e. 'bad, instead of :bad.  Is that syntax really to
> be allowed?  Can't find that in the spec...

Section 3.4.1:

   "A keyword-name can be any symbol, but by convention is normally
    a keyword[1]; all standardized functions follow that convention."

(the reason being that keywords are self-evaluating, so you don't have
to quote them at the call.)

See 3.4.1.6 for an example.

> Down to 328 failures from 441 with this patch.

I've been boosting it above 441 with recent test additions (460
this morning with my local copy).  Sorry to make the debugging seem
Sisyphean, but at least the test suite will not grow forever.

BTW, the test suite has caused some action over in the OpenMCL
world:  http://openmcl.clozure.com/Distributions/openmcl-release-0.13.3.html

        Paul