Bug#398373: RC class bug, dataloss grade, No 398373

[Josselin Mouette]

Le mardi 28 novembre 2006 à 12:38 +0100, Andreas Barth a écrit :
> * Josselin Mouette (joss at debian.org) [061128 12:36]:
> > Le mardi 28 novembre 2006 à 12:30 +0100, Andreas Barth a écrit :
> > > > FWIW, fixing this bug requires changes in the kernel.
> > > 
> > > Why that? AFAICR, umount must not return before not everything is
> > > written down.
> > 
> > I don't think this is always the case.
> > 
> > Of course, if the applet is saying things are OK while umount is still
> > running, this can be fixed in the applet, but I wonder whether this
> > would be enough.
> 
> I wonder too. But the current state is definitly dangerous, but - many
> investigations start with not knowing enough.

A first look at the code shows the callback called when umount has
completed is only used for displaying errors, not successful operations.
There's definitely something wrong here.
-- 
Josselin Mouette                /\./\

"Do you have any more insane proposals for me?"