Re: Session saving & 2.26.0

[Matthias Clasen <matthias clasen gmail com>]

On Wed, Mar 25, 2009 at 8:35 AM, Vincent Untz <vuntz gnome org> wrote:
> Le mardi 24 mars 2009, à 23:24 -0400, Matthias Clasen a écrit :
>> On Mon, Mar 16, 2009 at 10:20 AM, Vincent Untz <vuntz gnome org> wrote:
>> > Le lundi 16 mars 2009, à 10:36 +0000, Lucas Rocha a écrit :
>> >> Waiting for 2.26.1 and disabling session saving on 2.26.0 sounds like
>> >> the sanest option.
>> >
>> > We have a deal, then.
>>
>> I see that you have committed all your changes without further
>> discussion to the stable branch now. I don't think that changes of
>> this magnitude are compatible with our established policy for stable
>> branches.
>
> It was the plan that was agreed -- try to fix things for 2.26.1. If it
> turns out all this is broken, we can still revert things, or
> retroactively branch for 2.26.
>
>> Anyway, I'd like to thank you for breaking up the huge patch
>> into a series of reviewable commits. Looking over them, I can accept
>> almost all of them as  bug fixes of some sort, but this one stands out
>> as breaking the new gnome-session design:
>>
>> 2009-03-25  Vincent Untz  <vuntz {at} gnome.org>
>>
>>       Allow an interacting app to cancel the logout.
>>
>>
>> Can you justify that change ?
>
> Sure.
>
> This is basically only happening when a client is done with interacting
> during logout; note that an interacting client creates a JIT
> logout inhibitor.
>
> Basically, you already have the possibility to cancel the logout from
> the inhibit dialog (so I'm not quite sure why you're saying this breaks
> the current design). It turns out that some clients also display a
> button to cancel logout when they're interacting (gedit, for example).
>
> We could say that it's useless to support cancelling both via the
> inhibit dialog and via the clients interaction dialogs (this would of
> course ignore the fact that we don't control if apps are displaying a
> button to cancel logout -- ie, XSMP compatibility).

It is a big difference if you allow a user to cancel the logout or if
you let a random misbehaving client block logout. If you read Jons
design at http://live.gnome.org/SessionManagement/GnomeSession it has
a single sentence that is marked in bold (presumably because it is an
important part of the design):

    It should be stressed that, even when not forced, *clients should
not assume that
    they will have the ability to block logout or shutdown*.

You just broke this by giving clients that ability back.