bug#24372: 25.1.50; After losing focus, cursor is hidden when moving point

[Eli Zaretskii <eliz@gnu.org>]

> From: Philipp Stephani <p.stephani2@gmail.com>
> Date: Sun, 11 Sep 2016 17:37:36 +0000
> Cc: 24372@debbugs.gnu.org
> 
>  Using the maximum of blink-cursor-delay and blink-cursor-interval
>  effectively removes the user's ability of controlling the delay before
>  the cursor starts blinking when Emacs becomes idle, doesn't it?
> 
> Yes. I tried to figure out what the intention of blink-cursor-delay was, but couldn't find anything (the commit
> where it was introduced doesn't provide any motivation). I don't think that blink-cursor-delay <
> blink-cursor-interval ever makes sense. The other way round (delay > interval) is useful if you want to keep the
> cursor visible for a bit longer after a command.

I don't understand the nature of your difficulty.  blink-cursor-delay
is obviously how soon the user want the cursor to start blinking after
Emacs becomes idle.  That is orthogonal to the blink interval, which
is in effect once the blinking begins.

>  Plus, if the user sets the interval to a very small value, we have the
>  same problem again.
> 
> True, but so far that hasn't happened. Also it seems less surprising: if you increase the frequency, the cursor
> blinks faster, as expected.

The problem is not with a faster blinking, the problem is with the
cursor almost invisible.  E.g., try to set blink-cursor-delay to a
very small, but non-zero value with today's sources, and then repeat
your recipe, with and without the focus-out/in dance.

>  How about a variant of this below? It uses a fixed limitation from
>  below on the delay, but only for the first blink. (The value 0.2 was
>  found by experimentation, not sure if we need to add yet another
>  defcustom for that.)
> 
> I don't think we should introduce magic numbers or further customization options.

It solves the problem, doesn't it?  I don't mind very much if it were
a defcustom, I just think no one would want to change it.

>  > I've attached another patch with the change I have in mind.
> 
>  This has a disadvantage of creating a new timer object each time,
>  which I think we'd like to avoid: too much consing. (Also, don't you
>  need to set the timer variable to nil when the timer is disabled?)
> 
> I don't understand - the patch doesn't create any additional timers, it only changes the initial delay of the
> idle-timer.

Each time blink-cursor--start-timer or blink-cursor--start-idle-timer
is called, they create a new timer, right?  And your patch makes us
call these functions each time blinking is started or ended, right?

> My patch is identical, except is uses blink-cursor-interval as lower bound. 

Of course.  That's why I said it's a minor variant.

There's another difference, though: in my patch we only limit the
first argument to run-with-timer/run-with-idle-timer, not the second.
So only the first blink cycle is affected.