Re: [Lxde-list] taskbar performance enhancement

[PCMan <pcm...@gm...>]

On Wed, Sep 14, 2011 at 8:18 AM, Rouslan Korneychuk <rou...@ms...>wrote:

>
> ________________________________
> > Date: Tue, 13 Sep 2011 21:28:23 +0800
> > Subject: Re: [Lxde-list] taskbar performance enhancement
> > From: pcm...@gm...
> > To: rou...@ms...
> > CC: lxd...@li...
> >
> > In theory, this might decrease usage a little.
> > However, is the reduction of resource usage visible to the users?
> > If this makes the program even harder to maintain, this isn't worth the
> > effort.
> > In addition, using customized widgets means we lose accessibility
> > support provided by standard gtk+ widgets. We already have this
> > problems in the launch bar plugin and it's hard to fix.
> > If you have 10 windows opened, there will be 10 GtkToggledButton in the
> > taskbar. Ten instances of GtkToggledButton actually cost several
> > kilobytes only. The most resource they used in the button is for the
> > icon, which is a bitmap. Creating a custom taskbar won't save any
> > resource from this part. You still need the icons anyways.
> > So, I guess this might not make that much differences. It's just my
> > guess and it's not base on profiling or benchmark. However, if you have
> > the time and ability, I'd suggest that you devote your time to fix
> > other parts to make it better. This should be more cost-effective.
> > Cheers! :-)
>
> There was something compelling about this little project, so even if
> it's not important, I'm quite interested in doing it (I'll pick
> something more useful next). I already wrote a custom widget that
> displays the task buttons and responds to mouse movement and clicks,
> and the code is not very complicated. As for accessibility, I'll use
> the ATK library and test it with Accerciser
> (http://live.gnome.org/Accerciser). I'm a perfectionist anyway, so I
> don't mind.
>
> If you have other compelling reasons to do so, then it will be great since
you may use it in other places as well. Having a customized widget can make
theming the taskbar easier and we can have greater flexibililty. Actually,
part of the taskbar is painted with cairo already. This is formerly done by
Marty Jack, who is no longer using lxde now.
Thank you for your interests in working on the panel, but I must admit that
it's currently in a poorly maintained status. The initial fork from fbpanel
is done by Fred and me. Then later Marty Jack did a lot of improvement.
However, none of us can maintain the panel at the moment. (I'm preparing for
my license examination now)

The current plan for the panel is actually a redesign and rewrite with gtk3.
Porting existing code to gtk3 takes as much effort as a rewrite. In
addition, there has been quite a lot long-standing bugs and design flaws in
lxpanel. Personally I'm interested in a cleaner rewrite with gtk3. So I'd
suggest that you make the taskbar stuff gtk3 compatible.

Besides, I'm planning to use libwnck rather than doing all the dirty X11
stuff ourselves. It's not a huge lib and has no other dependencies. This can
save us quite a lot of time and solve several unresolved problems in several
places.
If we don't use the taskbar widget provided by wnck, we can still use its
APIs to retrive window information required by our taskbar. This makes the
code much cleaner and more readable.
If possible, making the taskbar with gtk3 + wnck and even better, with atk
support is encouraged.
As no new features are currently being developed for lxpanel, I don't think
you will create compatibility issues. Just git clone and do your
experiments. We can pull it to our repo easily later if your code works
well.
Thanks!


> > Date: Tue, 13 Sep 2011 13:34:17 +0200
> > From: br...@bs...
> > To: rou...@ms...; lxd...@li...
> > Subject: Re: [Lxde-list] taskbar performance enhancement
> >
> > [...]
> >
> > But seriously; if you can do something (branched) and someone can build
> > and test and do measures I am all for it. Don't get over the top dirty
> > with this before someone confirms that they can work with you.
>
> When I said I would need a little guidance on compatability, all I
> really meant was somebody on this mailing list would briefly explain
> how to do it. e.g.: Can I target a different version by setting a
> macro or do I need to download headers from an older version, and
> what's the oldest version I need to support?
>
>
>