bug#3984:

[Stefan Monnier]

> Oh, that's actually not necessary. I put it in so I could re-run the test
> without restarting emacs or manually resetting the advice, and I forgot to
> remove it.

OK, removed.

>> Since it's only used once, I don't think it's worth defining a macro for it.
> Probably true. I wasn't sure how much I would end up using it, and it seemed
> like a potentially useful macro to have lying around when testing advice
> that might break core functions of emacs.

Makes sense as well, but let's postpone putting it in until we do end
up using it repeatedly.

I installed some tests based on your patch, as well as my quick fix in
subr.el.

> That would certainly work, assuming that subr.el is always loaded before
> nadvice.el (so call-interactively could not possibly be advised yet), which
> seems likely to be true since subr.el contains so many core functions.

Since subr.el is both preloaded and loaded very early, that should be
safe, yes.

> Since this bug report is just about advising call-interactively, which we
> now have a fix for, should I open a separate bug about the around advice?

Sure.  Note also that in the tests I installed, I added one that
demonstrates a problem with :filter-args as well (because :filter-args
uses `funcall' rather than `apply', so it doesn't match the pattern
expected by advice--called-interactively-skip).

By the way, I don't see your name in our copyright assignment database.
I installed your tests under "tiny change", but if you intend to
contribute more, we'd need you to sign some paperwork.  Would you be
willing to do that?


        Stefan