Rasmus <rasmus@gmx.us> wrote:
> Kyle Meyer <kyle@kyleam.com> writes:
>>     ** TODO h
>>     DEADLINE: <2015-07-13 Mon +3w>
>>
>> and I run org-todo with a numeric prefix of -1, the repeater is canceled
>> by changing it to +0w:
>
> I see.  Shouldn't org-todo rather remove the repeater?

That seems fine to me.  I wonder if there was a particular reason the
original code set it to 0 instead.

>>        (+ daynr 1000)))
>> -   ((and daynr (string-match "\\+[0-9]+[hdwmy]" s))
>> +   ((and daynr (not (string-match "\\+0[hdwmy]" s))
>> +	 (string-match "\\+[0-9]+[hdwmy]" s))
>
> for no particular reason I prefer:
>
>    (and daynr (string-match "\\+\\([0-9]+\\)[hdwmy]" s)
> 	 (> (string-to-number (match-string 1)) 0))

I do too (though I think s should be added to the match-string call).

>> -	  (setq dn (string-to-number (match-string 1 change))
>> -		dw (cdr (assoc (match-string 2 change) a1)))
>> +      (when (string-match "\\(\\+[0-9]+\\)\\([hdwmy]\\)" change)
>> +	(setq dn (string-to-number (match-string 1 change))
>> +	      dw (cdr (assoc (match-string 2 change) a1))))
>> +      (unless (wholenump dn)
>
> Shouldn'it it also test that dn > 0?

Doh, it should.

I've attached an updated patch.  Thanks for your comments.