Re: Selecting tooltip frames considered harmful

[martin rudalics]

>> The obvious candidate here is the minibuffer window of tip_last_frame
>> (making sure the tooltip frame gets deleted when tip_last_frame is
>> deleted).
>
> I was rather thinking of the selected-window at the time the tooltip
> frame is created

A minibuffer or echo area window is specified frame-wise and is found
in the minibuffer_window slot of each frame.  So you probably mean the
mini window for the frame of the window selected at the time the
tooltip frame is created.  That window is in fact the mini window of
tip_last_frame.

> (plus some way to re-set the "transient-for" field so
> the same tooltip-frame can be reused?).

The same tooltip-frame (and its buffer and window) can be already
reused.  So what does the "transient-for" field stand for?

> First we need to decide what's The Right Thing, and I think that keeping
> track of the "parent" frame/window is the right thing.
>
> Once we agree on this, we can think about how best to
> implement/approximate it.

We currently refuse to delete a frame when its minibuffer window
serves as the minibuffer window of another frame.  Since this sounds
like a bad idea in the context of tooltips, we probably would have to
delete a tooltip frame before deleting the frame whose minibuffer
serves as the surrogate for the tooltip.

So yes, we have to agree on what TRT is.  Either insist on all frames
having a minibuffer frame (or producing one on demand) or disallow
selecting frames which don't.

martin