[LightDM] Fwd: Is it time to re-factor yet :^)

[Bryce Harrington]

On Wed, May 18, 2011 at 01:45:26PM -0400, j. apple muncy wrote:
> On Mon, May 16, 2011 at 11:56 AM, David Edmundson ***************
> > On Mon, May 16, 2011 at 4:20 PM, j. apple muncy <j.apple.muncy at gmail.com>wrote:
> >> Or how to avoid the whirlwind.
> >>
> >> I observe the coding of lighdm is well under way and I'm thinking the
> >> pressure to meet the releases of 11.11 is probably going to build without
> >> bounds.
> >>
> >> Is there time now to consider the inclusion of some of the more offbeat
> >> use cases?
> >> 4: No Dbus.
> >
> > When would "no dbus" be a real use case?
> >
> That would be for fail-safe mode when D-bus is broken. Not so very long ago
> I loaded up a make shift machine with 10.10 but about thitry seconds into a
> session it would dump back to the gdm screen. As nearly as I could
> determine, D-bus was not working as advertised.

There are likely a proliferation of different failure use cases, perhaps
it would be good to justify use cases via quantity of bug reports?

How many bug reports are open about dbus failing during or prior to login?

> But my point is that if LightDM is going to be the default DM, care must be
> taken to handle lots of different use cases.
> 
> >> 5: No X until after login.
> >
> > If you've already logged in, you don't need a login prompt for anything.
> > What would LightDM (or GDM/KDM/XDM) gain a user over simply typing startx
> > themselves? They're clearly the sort of user that doesn't like doing things
> > the easy way, maybe this software simply isn't for them.
> >
> >
> IMHO down thought the years of Linux installs, the single most difficult
> gotcha is when X fails to start. Sure, LightDM could just do the traditional
> thing and bailout to getty.

Sounds like this can be generalized into the "X is failing to start" use
case, which is legitimate (but thankfully exceedingly rare).

I'm planning on tending to this particular use case, and will work with
robert on it.  I think we can take "no dbus" as a restriction for
ensuring this mode works.  I don't plan on adding any support for
troubleshooting dbus failures but at least it will display error
messages from Xorg.0.log, so at least the user will be made aware of the
cause of the failure.

And, the failsafe-x session would be the right framework for adding
tools to repair/restart dbus, if it's a problem users run into
sufficiently often, and/or if someone would like to add this support.

> I look forward to Robert's return from UDS and the hot info about what is
> expected of LightDM for 11.10.

Sounds like Robert was hit with the Ubuflu in the final days of UDS so
he may need a bit of extra time for recovery.  (Get well soon Robert!)

Bryce