[asdf-install-devel] Patches - unit test improvements, clarifying refactorings

[Dan Muller]

Dan Muller keeps talking to himself:
> Dan Muller wrote:
>
> I've updated my "submitted" repository to include all of unstable,
> plus a conflict resolution patch for test-asdf-install.lisp.
>
> This made my unit tests on Linux work a bit better; but I've got new
> failures on Windows that I have to investigate. I figured it was
> worthwhile to get make the conflict resolution patch available in the
> meantime, though.

I have tracked the causes of the failures on Windows down and pushed
two more patches to submitted. This brings me back to where I was
before, with one unit test that errors out (direct-install : test-1),
apparently because I don't have Gary's key marked as fully
trusted. (Is that a prerequisite for passing some of the tests?)

I still have five unit test errors with SBCL on Debian. Some of those
seem to be due to an attempt to delete a non-empty directory. But
that's better than I was doing before in this environment.

Here are the patch descriptions, plus the description for the conflict
resolution patch:

Wed Dec 26 21:56:55 Eastern Standard Time 2007  Dan Muller<pikdj2002 at sneakemail.com>
  * Fixed PATHNAME-NAME+TYPE in TEST-ASDF-INSTALL to account for :UNSPECIFIC pathname-type.

Wed Dec 26 20:57:25 Eastern Standard Time 2007  Dan Muller<pikdj2002 at sneakemail.com>
  * Removed binding of *DEFAULT-PATHNAME-DEFAULTS* in INSTALL-PACKAGE.

  In LispWorks Personal 5.0.1, the call to DIRECTORY in INSTALL-PACKAGE
  sometimes returns NIL when it shouldn't. The problem seems to related
  to the binding and/or use of *DEFAULT-PATHNAME-DEFAULTS*. On
  examination, it seems that binding this is really pointless anyway,
  since it was also being passed explicitly to MAKE-PATHNAME. So I
  replaced it with a lexical variable. That solved the problem that I
  was encountering.


Wed Dec 26 19:45:22 Eastern Standard Time 2007  Dan Muller <pikdj2002 at sneakemail.com>
  * Resolving conflicts in test suites between DanM's changes and unstable.

  These occurred due to a somewhat stale repository.


-- 
Dan Muller