Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> I meant featurep, but since cl-assert is autoloaded, that shouldn't be
>
> AFAICT it's not autoloaded.

You're right, emacs -Q followed by (fboundp 'cl-assert) returns nil.

> I think the confusing comes from help-fns.el which does (require
> 'cl-lib), so if you do `C-h f cl-assert RET` it will indeed tell you
> it's autoloaded, but it's just because `C-h f` ended up requiring
> cl-lib.

No, the confusion comes from looking at the source of cl-macs.el, in
which cl-assert is given an ;;;###autoload cookie.  I see now that there
are some comments in cl-lib.el that suggest loading cl-loaddefs.el
involves some trickery, so I guess I don't get to eat the cake here.

In that case, I don't see how Eli's suggestion to predicate the
cl-assert call on (featurep 'cl-lib) will help, as feauturep is not
going to load the library for us.  Do we want the following instead?