performance regression caused by deskbar-applet default inclusion

Jeff web.kiddo at
Thu Sep 27 19:35:17 BST 2007

Darn, sorry for replying late, I forgot to check my mailing list
accounts :(

Le mardi 25 septembre 2007 à 09:57 +0200, Sebastien Bacher a écrit : 

> Le lundi 24 septembre 2007 à 17:32 -0400, Jeff a écrit :
> Could you give detail on how you mesured those 15 seconds ? Is that
> using a gutsy CD or on an installed version? GNOME should take less than
> 10 seconds to start on a such configuration

This is measured using a stopwatch (my simple wrist watch), on a fully
installed and up-to-date ubuntu gutsy desktop using gnome. Compiz is not
enabled (because it eats 15 secs login time too!).

My session started items are: 

      * the system-config-printing notification area icon 
      * gnome power manager 
      * gnome volume manager 
      * network manager 
      * tracker (however, I set it to start only after 50 seconds; and I
        "feel" it starting [scratching the drive] waaaay after the
        desktop is loaded, deskbar or not) 
      * update notifier

My panel was, at the moment of that test, pretty clean (and did not
change between iterations). From what I remember, it had: 

      * gnome menu bar 
      * window list applet 
      * notification area 
      * 4 launchers 
      * sensors-applet

To make the benchmark somewhat rigorous, I make a cold start (that means
powering off the machine entirely) between the two tries. Starting with
the deskbar applet vs starting without that applet on my panel makes a
15 seconds difference.

Where is my federico-promised 3 seconds gnome login time :) ?

Here is what I consider "loaded" and what I consider "not ready": 

      * all panel applets must be loaded 
      * desktop must be loaded (in my case, I don't even have nautilus
        drawing the desktop so...) 
      * the hard drive LED must have stopped flashing (it must be *off*)

Unless there were some very serious performance optimizations between
deskbar 2.19.92 and 2.20, I would believe that my 13-15 seconds
(approximative) login time difference still applies.

Also, my laptop is not using autologin, so I start my stopwatch at the
*precise* moment when I hit the "enter" key after entering the password.
That is my starting point in order to be somewhat methodical, otherwise
you can't have any precision with autologin (I think).
-------------- next part --------------
An HTML attachment was scrubbed...

More information about the ubuntu-desktop mailing list