On Mon, Feb 19, 2001 at 11:40:13PM +0100, Martin Atzmueller wrote:
> I've been hacking on PCL these days.
> It mainly was about fixing bugs, and finding out how PCL worked.
> Then I went along your "FIXME" notes in the code, and cleaned up
> some portions of the code.
> I mainly removed non ANSI hacks, hacks that are not needed any more,
> ugly compiler hacks, and I enabled some portions, that were
> commented out.
> - non ANSI: (setf <method>) handling,
> - non ANSI: *defgeneric-times*, *defmethod-times*, *defclass-times*
> - make-top-level-form replaced by "eval-when" or simply removed, when
> - and lots more ...
> I also generalized a lot, so that one function is used for one purpose,
> instead of two (instance-ref vs %instance-ref), and I moved some stuff
> to internal-packages. I tried to use "internal"-functions as far as that
> was possible ("once-only" ...).
> There were also some functions that were not in "their" package, so I
> tried to move them there.
> Finally I moved some functionality concerning the compiler to a seperate
> file, since one file (low.lisp) had two "IN-PACKAGE"-forms.
> So, after all this, I think this should be tested, and since it works as
> far as I can tell, it probably should go into the main source.
> I have used PVE's ansi-test suite, specifically the "clos-test", and the
> "new PCL" gives the same "CORRECT"'s (and the same errors .... :(
The ansi-test suite is great, but I wish it would identify tests, and
failures, in such a way that we could maintain a list of things that
are known to fail, so that we could automatically skip over
corresponding failure messages. I really like being able to run tests
automatically, with a simple summary of whether I got the expected
result, and I haven't been able to figure out how to do this with the
ansi-test stuff (short of fixing every last bug in SBCL, which might
take a while). In the meantime, before we've fixed every last bug,
it's valuable to know whether we're introducing new bugs, and
ansi-test would be a powerful tool to catch them.
> I could use some feedback before I work further on this,
> so any feedback is appreciated. :)
All the stuff summarized above sounds good to me, as you might guess
if you've been reading my FIXME notes. My only reservation is that
24kB (gzipped!) of changes is likely to be quite a lot to grok all at
once. But I'll try to give a more detailed response sometime this week.
William Harold Newman <william.newman@...>
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C B9 25 FB EE E0 C3 E5 7C