Re: emacs-30 b585826a65e: ; * lisp/files.el (require-with-check): Fix doc string and error text.

[Eli Zaretskii]

> Date: Wed, 14 Aug 2024 05:56:21 +0000
> From: Pip Cet <pipcet@protonmail.com>
> Cc: emacs-devel@gnu.org, monnier@iro.umontreal.ca
> 
> "Eli Zaretskii" <eliz@gnu.org> writes:
> 
> >> +         ((and fn (memq feature features))
> >> +          (funcall (if noerror #'warn #'error)
> >> +                   "Feature `%S' is now provided by a different file %s"
> >> +                   feature fn))
> >> +         ((funcall (if noerror #'warn #'error)
> >> +                   "Could not load file %s to provide feature `%S'"
> >> +                   (or filename (symbol-name feature))
> >> +                   feature)))))
> >
> > The second message is not accurate, AFAIU.
> 
> Can you provide an example for a case in which the message is emitted
> but is not accurate?

When fn is nil.

> > If NOEROR is equal to
> > 'reload', and the file for FEATURE could not be found,
> > require-with-check will signal an error from here:
> >
> >   (let ((lh load-history)
> >         (res (require feature filename (if (eq noerror 'reload) nil 
> > noerror))))
> 
> Correct, so we would never reach the code emitting the new message.

But the message text mainly describes this situation: "Could not load
file".

> > Also, if FILENAME was found, but it doesn't provide FEATURE, the
> > message text again doesn't provide accurate diagnostic.
> 
> If FILENAME exists and is loadable, but doesn't provide FEATURE,
> 'require' will throw an error (even if NOERROR is t).  So, again, we'd
> never reach the code emitting the new message.

And again, the error message will not describe the situation.

> > So I think we need 3 different messages, not 2.
> 
> I don't understand what the third message would be.

For starters, that no FILENAME could be found for FEATURE.