bug#20241: 25.0.50; `setq' with only one argument

[Alan Mackenzie <acm@muc.de>]

Hello, Glenn.

On Tue, Nov 24, 2015 at 12:38:18PM -0500, Glenn Morris wrote:

> I'm assuming that any uses of this form were intentional (or even if
> not intentional, were doing the right thing), rather than errors.

That wasn't the case.  There were 5 deliberate uses of `setq' without
the nil, there was one error in bytecomp.el, which had the effect of
suppressing warnings given out by the byte compiler.

> (I have no actual data - do you?) So making such code stop working
> overnight seems too harsh to me.

I have data, yes, see above.  The point is, that error in bytecomp.el
could have been anywhere, causing random breakage.  I only noticed it by
accident.

I don't think the value of allowing an implicit, but undocumented, nil
at the end of a `setq' outweighs the danger of accidentally setting a
variable to nil.  There were, after all, only five uses of this
"feature" in the entire Emacs code base.  And one of these was in
.../obsolete.

-- 
Alan Mackenzie (Nuremberg, Germany).