Re: Improving Nautilus startup time

[michael meeks <michael meeks novell com>]

On Wed, 2006-03-08 at 18:33 -0600, Federico Mena Quintero wrote:
> I wrote a little program to extract a timeline of Nautilus's startup
> from an strace log.  The first timeline is here:
> 
> http://primates.ximian.com/~federico/news-2006-03.html#login-time-1

	So - even on the latest traces - it seems the time to init & startup
gnome-vfs-deamon is not insignificant.

	Is that really necessary at startup / bootstrap ? can we get to render
the desktop background & a few stock icons before doing this ?

	Seemingly at least the 1st set of calls:

	/* Watch for volume unmounts so we can close open windows */
	g_signal_connect_object (gnome_vfs_get_volume_monitor (), "volume_unmounted",
				 G_CALLBACK (volume_unmounted_callback), application, 0);
	g_signal_connect_object (gnome_vfs_get_volume_monitor (), "volume_pre_unmount",
				 G_CALLBACK (volume_unmounted_callback), application, 0);

	These 2 can trivially be delayed (surely) until we actually have some
open windows [ ignoring the 'root' window ] to close ? ;-)

	g_signal_connect_object (gnome_vfs_get_volume_monitor (), "volume_mounted",
				 G_CALLBACK (volume_mounted_callback), application, 0);

	This is more tricky I guess; but - presumably - (again ;-) at some
stage during bootstrap (hopefully at idle?) we query the system for the
currently mounted volumes to display pretty icons for them on the
desktop; surely only after that is it worth connecting this mount
callback ? [ or does this handle hot-plug during the login process
itself ? (surely somewhat unlikely) ].

	Of course - perhaps there are other vfs_get_volume_monitor calls
lurking that are not deferred until necessary.

	Federico - have you tried just hacking those 3 connects out to catch
any others ?

	Regards,

		Michael.

-- 
 michael meeks novell com  <><, Pseudo Engineer, itinerant idiot