Re: gconf backend

[Havoc Pennington <hp redhat com>]

On Sat, 2003-09-27 at 12:14, Daniel Veillard wrote:
> 
>   Okay, the small rant apart, do you suggest I go dig in to the gconf
> parsing code, to try to increase parsing speed ? I could go there once
> 2.6.0 is finished.

You are more than welcome to do so - clearly nobody had ever
memory-profiled gconfd at all or it would have been immediately obvious
we should free the DOM trees after loading instead of keeping them
around. ;-) So if you tried profiling it you may well find interesting
things to improve.

What you should profile is the new backend, I guess someone would have
to convert it to libxml first; the main point of the new backend is to
improve the overall code quality and sanity. I think you'll find the old
backend's code to be incomprehensible. I certainly don't understand it.
Mark may have figured it out, I'm not sure. ;-)

>   Do we have some oprofile measures indicating how much gconf initialization
> costs when starting a session ? 

Not that I know of, but Will Cohen may even be interested in helping to
create the profile; he likes to have sample cases of using oprofile.

Havoc