bug#12833: **: Re: bug#12833: 24.2.50; Avoid questions asked in the echo area to be overridden by messages

[Sebastien Vauban]

Hi Stefan,

Stefan Monnier wrote:
>> Schedule a date from that item with `C-c C-s': the prompt for the date gets
>> hidden (for something like 5 seconds) by the "require" message (though,
>> that happened before displaying the prompt).
>
>> This is reproducible ever and ever, in the same session.
>
> It shouldn't be 5 seconds, but more like 2 seconds (or rather "whatever time
> it takes to display the messages plus setup the minibuffer plus 2s").
>
> IOW this is `minibuffer-message-timeout'.

Just checked that unknown var (for me): yep, equals 2. The true reality (now
that I'm chronometering it) is as well that the messages are more or less
displayed for 2 seconds -- but they seemed like 5 looong seconds.

> What behavior would you like to see instead?

IIUC, the require (of some package) is done before asking whatever question in
the minibuffer. So, for me, the same would apply as in other cases...

    If I do 2 requires in a row (package "A", then package "B"), the message
    telling I required the first one ("A") will certainly not be visible in
    the echo area, as immediately after, there is another message being
    displayed, saying that "B" just got required.

What does that change in case of interactive question?

    If I do 1 require before asking a question to the user, the message about
    that require should hardly be visible (if visible at all -- because
    actions happen so quickly). The question should use the space of the
    information message, point.

In the given example here above, things are happening in the function
`org-read-date':

--8<---------------cut here---------------start------------->8---
(defun org-read-date (...)
  "Read a date..."
  (require 'parse-time)
  ...
  (read-string prompt default-input 'org-read-date-history timestr)
  ...)
--8<---------------cut here---------------end--------------->8---

The require of `parse-time' does display a message in the minibuffer (echo
area, to be accurate). When executing `read-string', whatever displayed
message should disappear and be replaced by the prompt.

The prompt, IMO, should never be overridden by a reminiscent message.

Am I exposing clearly my point of view (be it shared or not)?

Best regards,
  Seb

-- 
Sebastien Vauban