[vlc-devel] [vlc-commits] xcb: support =?utf-8?q?for=09WM=5FDELETE=5FWINDOW?=(WM_PROTOCOLS)

[Rémi Denis-Courmont]

Le samedi 29 janvier 2011 12:16:46 brezhoneg1, vous avez écrit :
> Yes, as the default window provider, it ends up being used in very
> different situations, and therefore it's not obvious to provide  a
> generic  "close feature" that suits all needs.
> 
> For the time being, do you suggest a full revert of these patches (Linux
> and Windows)  or trying to find a better tradeoff ?

I find that closing the whole player on closing the video window is really 
confusing if anything. (MDI) Applications should not exit just because one of 
their windows is closed. Only the main UI window should behave that way.

It's all the more true as VLC has been working that way for ten years, and we 
are only likely to confuse users if we change. I myself have had a few 
"unintended exit" accidents with the new built-in playlist: I mixed the main 
window containing the playlist for the detached playlist dialog... I think it 
would be worse with the video.

This is also obviously inappropriate for LibVLC.

So it all comes down to corner cases like the wall filters, duplicate video, 
and headless VLC instances. For the filters and duplication cases, I find that 
leaving the video on is actually a feature, more than a bug...
For non-graphical interfaces, I think it's more consistent to keep the same 
behaviour: if you want to quit, quit the interface, e.g. Ctrl+C (all cases), 
or Ctrl+D specifically for RC.

> Actually, the missing brick is the ability for a vout_window_t object to
> pass information back to the vout_display_t object, since a
> vout_display_SendEventClose is already available. (Though it also ends
> up in a playlist_Stop with a FIXME ..., problem is passed and what to do
> remains unclear)

That was the historical behaviour from before VLM and LibVLC: stop the 
playlist. But it would need to be cleaned up so that it accomodates the 
different "input managers" nowadays.

-- 
Rémi Denis-Courmont
http://www.remlab.net/
http://fi.linkedin.com/in/remidenis