Re: Desktop slowness issue - freeze break request

Le mercredi 31 août 2005 à 08:36 -0600, Elijah Newren a écrit :
> On 8/31/05, Murray Cumming <murrayc murrayc com> wrote:
>         [snip]
>         > I argue in the bug why the omission of these redraws is
>         safe, but even
>         > if you don't believe that such a change can at most cause
>         redraw issues
>         > which are not very serious.
>         >
>         > This patch probably doesn't remove all the slowdowns caused
>         by this 
>         > apparent X slowness issue, but it at least makes it better
>         by limiting
>         > its scope. And it should help everyone, since these desktop
>         redraws are
>         > just generally a waste of time, even on systems where its
>         not very slow. 
>         >
>         > Is this ok to commit?
>         [snip]
>         This doesn't seem to be a regression, so maybe it can wait for
>         2.12.1. I
>         also don't see anybody confirming that it fixes the issue for
>         them.
> In my reading of the bug report, it looks like it is a regression
> caused by an Xorg bug triggered by cairo and this patch is just an
> attempted workaround.  Did I misread?
> I don't think this patch would have any unintended side-effects so it
> should be safe, but I'd be rather surprised if it actually helped.
> There was a comment in the bug that the problem was painful for
> focus-follows-mouse, but if they're using Metacity, the desktop window
> is never focused in any focus mode unless the user clicks on it:
>     case EnterNotify:
>        <snip>
>           switch (meta_prefs_get_focus_mode ())
>             {
>             case META_FOCUS_MODE_SLOPPY:
>             case META_FOCUS_MODE_MOUSE:
>               if (window->type != META_WINDOW_DOCK &&
>                   window->type != META_WINDOW_DESKTOP)
>                 {
>                   meta_topic (META_DEBUG_FOCUS,
>                               "Focusing %s due to enter notify
> Now, perhaps they're using a different window manager, or perhaps
> they're clicking on the desktop enough for this to cause problems, but
> the Mandriva bug Frederic linked to in that report suggested that
> mapping any new windows was causing problems, which means that closing
> windows is likely to have the same problem, and that's one you
> probably can't avoid--it may, in fact, be the problem.

Yes, it is really slower to move or close windows. Even doing right
click on background takes 3 or 4 seconds to popup.

> So, although the patch looks safe enough to me, I'd really rather see
> if it actually helped anyone first.

I've integrated the patch in Mandriva cooker.

I'll be able to test it on my home system tonight and report tomorrow.

Frederic Crozat <fcrozat mandriva com>

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]