Re: [gamin] gamin /media/* polling and inotify

[Brandon Hale <brandon smarterits com>]

On Mon, 2005-03-07 at 11:25 -0500, John McCutchan wrote:
> On Mon, 2005-03-07 at 10:43 -0500, Brandon Hale wrote:
> > On Mon, 2005-03-07 at 10:23 -0500, John McCutchan wrote:
> > > On Mon, 2005-03-07 at 09:30 -0500, Brandon Hale wrote:
> > > > In my testing, it appears that the recent polling changes in gamin have
> > > > solved the issues with /media/* and /mnt/* using dnotify. However, using
> > > > an inotify .19/.20 kernel, it appears that polling is not used and we
> > > > see the same bugs as before. Clues would be appreciated, I'll follow up
> > > > soon with testgam results.
> > > 
> > > 
> > > inotify should never block unmount, are you sure it isn't something
> > > else?
> > 
> > The problems I experience without falling back to polling aren't related
> > to unmount blockage. A few Gnome bits watch this directory (via gamin,
> > fam) for new devices. In particular, drivemount-applet and nautilus show
> > new devices as they are discovered in /media.  This now works with gamin
> > and dnotify. However, it still fails to work as expected using inotify.
> > As no one here has an immediate clue, I intend to look at the dnotify
> > vs. inotify code and see if perhaps the polling fallback for /media/*
> > and /mnt/* was not copied to the inotify backend.
> 
> I see. Well FYI the inotify backend is just a proof of concept. It is
> not feature complete and does not make use of the poll backend in any
> way.

Hm the calls to the poll are in functions that don't appear to have
obvious counterparts in gam_inotify.c, this is unfortunately beyond my
current skills to produce a patch. Thanks all for the useful
information.

Cheers,
Brandon Hale