Re: Remove double negation error messages.

[Jesper Louis Andersen]

Quoting Ralf Wildenhues (address@hidden):

> Thanks for reporting this.  I agree that double negation is suboptimal.
> However, since the upper bound is chosen on rather pragmatic grounds
> (we are pretty sure that five digits are portable), I'd rather not set
> it in stone.  In doubt we might raise the limit again.
> 
> How come you saw this?  Do you actually need higher limits?

Well, I was in the process of upgrading software for OpenBSD which still
has an older 1.5.10 libtool. This together with Gnome 2.10 using a 1000
CURRENT number got it to bomb (this was before the number increase). This
got me around to the error message, which I found suboptimal (1000 _is_ a
positive number, so I wondered why it bailed. I little grep later I had
the reason). 

This made me report the documentation error.

> How about this patch (against HEAD, other branches similarly)?

I have no problems about the patch. I would have preferred a separate
function to predicate if a number was positive (it is after all used a
number of places in the code), but how to crank this out of sh(1) so it 
is portable is beyond me at the moment (eating a character at a time and
checking it seems reasonable, but it might be slow). 

I do like the idea of warning the user that it is not the whole space of
positive numbers there are legal values. But we can come to a compromise,
where the double-negation has been killed. This was after all the worst
part of the error message.

-- 
jlouis