bug#37507: 26.3; Non-number value for `fill-column' causes wrong-type-arg error for `C-h f'

[Drew Adams <drew.adams@oracle.com>]

> > `fill-column' is supposed to have an integer value > 0.
> >
> > But if it has a non-numeric value then other functions
> > can barf, unless they watch out for that.
> 
> Shouldn't users just not set it to a non-numeric value then?

Should?  Yes.

Just?  I don't see why.

> I'm sure you can cause all sort of errors by setting
> things to invalid values.

And?

If this is only about users setting the value
then the answer is for them to always use
`set-fill-column'.  "Should [eom]."

But is there a reason why this shouldn't be protected
here?'

I wonder even whether there is a reason why
`current-column' shouldn't prevent that particular
error by, say, substituting 0 for a `fill-column'
value that's not an integer > 0.  I don't propose
that at this point (might break something that
depends on raising an error?), but I wonder if it
wouldn't be reasonable for `current-column' to do.