Re: Bug: Items with repeating timestamps don't appear in the agenda [6.36trans (release_6.36.509.g9e9b)]

[Bernt Hansen <bernt@norang.ca>]

Tassilo Horn <tassilo@member.fsf.org> writes:

> Hi again, Bernt!
>
> I did some debugging.

Thanks.  Sorry I've been tied up with work today.

> The reason that the repeating events are not shown is that
> `org-agenda-get-timestamps' doesn't return those repeater events,
> because when it iterates over all timestamps of a file, the test in
> the following `if' always returns t and so we are skipping.
>
> --8<---------------cut here---------------start------------->8---
>   (catch :skip
>     (and (org-at-date-range-p) (throw :skip nil))
>     (org-agenda-skip)
>     (if (and (match-end 1)
>          (not (= d1 (org-time-string-to-absolute
>                  (match-string 1) d1 nil
>                  org-agenda-repeating-timestamp-show-all))))
>         (throw :skip nil))
> --8<---------------cut here---------------end--------------->8---
>
> This happens, because `org-time-string-to-absolute' returns the first
> date (the literally written date), not the closest repeating date.
>
> For example, for "<2010-06-28 Mon 22:00 +1w>" (last week's Monday) that
> function returns 733951, although it should return 733958 (last week's
> Monday + 7 days = today), I guess.
>
> The underlined predicate looks suspicious to me:
>
> --8<---------------cut here---------------start------------->8---
> (defun org-time-string-to-absolute (s &optional daynr prefer show-all)
>   "Convert a time stamp to an absolute day number.
> If there is a specifyer for a cyclic time stamp, get the closest date to
> DAYNR.
> PREFER and SHOW-ALL are passed through to `org-closest-date'.
> the variable date is bound by the calendar when this is called."
>   (let ((today (calendar-absolute-from-gregorian (calendar-current-date))))
>     (cond
>      ((and daynr (string-match "\\`%%\\((.*)\\)" s))
>       (if (org-diary-sexp-entry (match-string 1 s) "" date)
>           daynr
>         (+ daynr 1000)))
>      ((and daynr (not (eq daynr today)) (string-match "\\+[0-9]+[dwmy]" s))
>                  ^^^^^^^^^^^^^^^^^^^^^^
>        (org-closest-date s (if (and (boundp 'daynr) (integerp daynr)) daynr
>                  (time-to-days (current-time))) (match-string 0 s)
>                  prefer show-all))
>       (t (time-to-days (apply 'encode-time (org-parse-time-string s)))))))
> --8<---------------cut here---------------end--------------->8---
>
> Why shouldn't that case trigger when daynr is today?  In that case, we
> don't even look at the timestamp s...

This change was my second attempt at fixing the scheduled display when a
weekly TODO entry is missed (ie it's over a week late already).  The
agenda display.  For the case where we're displaying the scheduled date
for today's agenda I wanted it to reflect how many days late the task is
instead of rolling over on the cyclic repeater.

I don't (yet) understand why your example works on my agenda but not
yours ...

Oh!  It's not that the future dates are missing it's the entry for
_today_ that is missing.  So yes my commit breaks this and I'll get it
reverted.

Thanks for reporting this issue!

Regards,
Bernt

>
> I deleted the whole underlined predicate, and then the agenda works
> again for me.
>
> Bye,
> Tassilo