Re: Still bugs in read-file-name completion on w32

[Mathias Dahl]

> > I think this is by design. Completion is there to help the user choose
> > a file, not to prevent them from choosing the wrong file. By default we
> > ignore backup files and known binary suffixes, but if the only match for
> > what the user has typed is a file that we would normally ignore, then we
> > happily complete to that, because ultimately the user knows best.

> Are you sure?
> I know this is the expected behavior of completion-ignored-extensions, but
> I didn't know this is the expected behavior of the `predicate' argument to
> read-file-name.

I agree, and the documentation string is quite clear:

...
If optional sixth arg predicate is non-nil, possible completions and
the resulting file name must satisfy (funcall predicate NAME).
...

*Must* satisfy.

/Mathias