CMU CL: #+cmu vs. #+cmucl

[Robert Goldman]

On 8/20/16 Aug 20 -11:55 AM, Elias Pipping wrote:
> So to summary what I’d like to ask:
>  - Is it known that tests fail on cmucl <20e?

TL;DR: I'm afraid not.  I am only testing on 20f right now.

I don't have a server farm (nor the time to manage one) and the most I
can manage is to run tests on a reasonably up-to-date set of lisp
implementations.

In practice, this generally means latest binaries available on most
platforms, although I usually test on a relatively up-to-date copy of
SBCL from git, and I have been testing on ECL and MKCL from source.

I am not competent to build lisp implementations on Windows, so for that
platform I am strictly limited to what I can download and run.

>  - What versions of cmucl is ASDF meant to support?

If Faré reports that the CMUCL maintainer prefers :cmucl feature, then I
propose we rule out anything that doesn't have it.  That would mean 20c
and forward.  We know that 20c and 20d have CLOS issues that we cannot
work around, so I would eliminate them, as well.

That would limit us to 20e and forward.

If there's a bug fix that would make 20c and 20d support better without
too much fuss, then I would be happy to merge it.  But I believe we
convinced ourselves that the CLOS issues could not be overcome.

What

>  - Should I turn checks for the :cmucl feature into ones for :cmu?

Sounds like "no," and we should replace old :CMU checks with :CMUCL.

Follow-on question: should we make ASDF issue a warning if it starts up
on older versions of CMUCL to inform the user that ASDF operations on
this implementation will be imperfect?

best,
r