[gamin] Re: gam_subscription_wants_event

[Daniel Veillard <veillard redhat com>]

On Tue, Jun 14, 2005 at 04:28:33PM +0100, Neal H. Walfield wrote:
> At Tue, 14 Jun 2005 11:25:39 -0400,
> Daniel Veillard wrote:
> >   the force option comes down from gam_server_emit_event. I can see it
> > called with both 1 and 0 in the code of gam_poll.c . There are many 
> > various cases, I'm pretty sure that if I did there was a reason, and
> > that reason still seems valid from a quick glance at the code.
> 
> Fair enough.  I am just trying to understand that reason :-)

  I don't have a map of all the transition possible for a monitored resources.
there is a number of possible switches:
   - monitored or not
   - exists or not
   - is of the expected type or not (dir vs. file)
   - busy or not
and there are more state informations (size, mtime, etc...)
The amount of transitions is fairly huge. There is no way I can have
a map there. Plus there is the additional complexity due to the indirection
forced by dnotify forcing to catch everything from directories.

That's why I'm so worried about casual changes which may just break 
the service for some kind of use patterns, and why I insists on running
the regression tests, python ones included before any commit.

Daniel

-- 
Daniel Veillard      | Red Hat Desktop team http://redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/