[PATCH 5/9] add a mode commandline option

[Ray Strode]

Hi,

On Tue, Feb 24, 2009 at 6:26 AM, Charlie Brej <plymouth at brej.org> wrote:
> william.jon.mccann at gmail.com wrote:
>> Initially supports modes: boot, shutdown.  This allows the
>> progress cache to be loaded from the appropriate file.
>
> A couple weeks back I had a bit of a hack around with the idea of showing
> plymouth on shutdown and although I could get it to do its stuff I was a bit
> scared of starting the splash before X had finished.
I think Jon got around this issue by not showing the splash until after X ends.

He posted some stuff here on the downstream Fedora bug:

https://bugzilla.redhat.com/show_bug.cgi?id=473237

Where he only starts plymouthd after prefdm has stopped.  We probably
want a better hand off at some point so there's no flicker, but this
should be fine until we have plymouth using libdrm directly and X on
top of drm modesetting in more hardware.

> Also I wasn't sure about
> the sequence of unmounting disks and the killall and trying to shutdown plymouth
> before it gets killed. Might be easiest to let plymouth be killed and let the
> shutdown duration file not be updated.
Right, we just ditch shutdown duration.  It's not clear it would be
that useful anyway, since it could only ever have some of the shutdown
messages in it.

I agree we should just let upstart kill plymouth.

> Do we want the splashes to be aware that they are in shutdown mode? What would
> they do different?
Maybe nothing.  I guess we can put the flexibility in place though if
they want to do something.

>  >no bool in structs please. Instead:
>  >uint32_t was_set : 1;
>  >It packs better that way.
> Just wondering, what happens when you ask for a pointer to a packed value?
you can't take the address of a bit field, so getting the pointer in
the first place would rely on some hack to get at the right offset in
the struct.  Then you'd get an int back with all the bits of the
bitfield in undefined order.

--Ray