Actually, I already [backported this change to abcl-1.1.x][r14349], mainly to keep any diffs in CLOS easy to perform from trunk.

Undoing [r14349][] won't help us ship asdf-2.27, so we should figure out what we should do to fix things.

The right way to fix this will be to implement an autoload mechanism for SETF macros, but that has proven to be taking more of my (mostly non-existent) time than I have.

Any suggestions on a hack that will load things correctly? I assume we can't just add the form to autoloads.lisp, as that will bring in CLOS on every startup. What about a (require 'documentation) at the end of clos.lisp? I'll go experiment with that idea...

@Fare: so definitely go ahead with shipping asdf-2.27, as ABCL's problems with ASDF are self-inflicted.

A: Because it messes up the order in which people normally read text.

Q: Why is top-posting such a bad idea?

A: Top-posting.

Q: What is the most annoying thing in email?

On Jan 24, 2013, at 9:30 AM, Rudolf Schlatte <rudi@constantly.at> wrote:

On Jan 24, 2013, at 07:31, Faré <fare@tunes.org> wrote:

Ahem, it works for me with ABCL 1.1.0.

Is it a bug in current ABCL?

Is there some workaround I could or should use in ASDF 2.26.143?

We autoload some functions to decrease startup time; I made documentation autoloaded on trunk, moving it out of clos.lisp (where it had no business being). This change is not currently in the next release (branch 1.1.x in svn), only on trunk.

The bug is that autoloading for a function foo currently does not work for forms of type (setf (foo x) bar), only for (foo x). Hence, a possible workaround is inserting something like

(documentation 'car) ;; fix (setf documentation) on abcl trunk

before trying to change docstrings, to ensure that (setf documentation) is available.

Cheers,

Rudi