Re: [bug] GNOME loses all settings following failure to resume from suspend

[Hugo Mills <hugo@carfax.org.uk>]

On Wed, Jan 05, 2022 at 09:39:21PM +0000, Hugo Mills wrote:
> On Wed, Jan 05, 2022 at 09:31:57PM +0000, Hugo Mills wrote:
> > On Wed, Jan 05, 2022 at 08:38:37PM +0000, Filipe Manana wrote:
> > > On Wed, Jan 5, 2022 at 6:34 PM Hugo Mills <hugo@carfax.org.uk> wrote:
> > > >
> > > >    Hi, Filipe,
> > > >
> > > > On Wed, Jan 05, 2022 at 06:04:38PM +0000, Filipe Manana wrote:
> > > > > I don't think I have a wiki account enabled, but I'll see if I get that
> > > > > updated soon.
> > > >
> > > >    If you can't (or don't want to), feel free to put the text you want
> > > > to replace it with here, and I'll update the wiki for you...
> > > 
> > > Hi Hugo,
> > > 
> > > That would be great.
> > > I don't have a concrete text, but as you are a native english speaker,
> > > a version from you would sound better :)
> > > 
> > > Perhaps just mention that as of kernel 3.17 (and maybe point to that
> > > commit too), the behaviour is no longer guaranteed, and we can end up
> > > getting a file of 0 bytes.
> > 
> >    I'd rather not reinforce the wrong usage with an example of it. :)
> > Better to document the correct usage...
> > 
> > > So an explicit fsync on the file is needed (just like ext4 and other
> > > filesystems).
> > 
> >    At what point in the process does the fsync need to be done?
> > Before/after/instead of the sync?
> 
>    Hmm. That doesn't make sense, of course (sorry, it's late, I've had
> a hard day). I'm guessing that the fsync needs to go after the write
> of the new data and before the rename. Is there any other constraint
> on what needs to be done to make this work safely?

   Right, I think I've got it. Ping me in the morning if it's not
correct.

   Hugo.

-- 
Hugo Mills             | Great films about cricket: Monster's No-Ball
hugo@... carfax.org.uk |
http://carfax.org.uk/  |
PGP: E2AB1DE4          |