[Gcl-devel] Additional tests

[Paul F. Dietz]

Some messages I sent to gcl-devel yesterday seem to have gotten lost,
so here's a recap.

I'm now seeing 144 failures out of 4367 tests.  New test
failures include:

  MACROLET is not handling &WHOLE parameters except at the top
  level.

  COMPILE has some minor interface problems, and also fails to
  respect object identity (up to EQL) as it is required to do.
  Getting the latter to work with gcl's C compilation strategy
  probably means generating C code with hardwired address constants
  when compiling anonymous lambdas.

  DESTRUCTURING-BIND is substantially nonfunctional -- it doesn't
  appear to recognize any lambda list keywords.

  Ordinary lambda list processing has a couple of problems (as
  illustrated by FLET.27 and FLET.30).  The argument
  :allow-other-keys nil is supposed to always be permitted (to
  functions with &key), but is being rejected.  Also, non-keyword
  keywords are not accepted as they should be.  I should extend
  these tests to LABELS, LAMBDA, and DEFUN.

In the future I will try to annotate failing tests with a comment
pointing back to the relevant section(s) of the CL spec.

I still need to go through the failing package tests.

        Paul