#782609 systemd-sysv,plymouth: user experience: plymouth animation "hangs"

Package:
plymouth
Source:
plymouth
Description:
boot animation, logger and I/O multiplexer
Submitter:
Simon Richter
Date:
2015-04-21 13:45:05 UTC
Severity:
normal
#782609#5
Date:
2015-04-14 18:50:30 UTC
From:
To:
Hi,

I've recently installed jessie on a netbook (Acer Aspire One 725), with
systemd as init and plymouth to hide the startup process from the user.

The boot process is about as quick as can be expected on this system, and
plymouth is stopped before the X server starts, also as expected, which
stops the animation. From this point, the X server requires a further 20
seconds to be in a state to display the login screen, most of this time is
spent waiting for socket-activated services to start.
animation presented during boot suddenly "hangs" for several seconds, and
then suddenly the login screen is shown.
intended, however the combination gives a bad result.

   Simon
#782609#10
Date:
2015-04-14 19:03:56 UTC
From:
To:
Am 14.04.2015 um 20:50 schrieb Simon Richter:

Can you elaborate why you filed this bug against systemd-sysv?
#782609#15
Date:
2015-04-14 19:20:05 UTC
From:
To:
Hi,

Basically, because this does not happen with sysvinit -- the boot
process takes longer there, but the X server startup is instant as the
required services are already running, so the boot process never
appears to hang.

I've filed the bug against both plymouth and systemd-sysv, because the
problematic behaviour is in the interaction between both -- as said,
each component is fully correct, but the end result is suboptimal. In
theory, xserver-xorg could also be included in the list.

   Simon
#782609#20
Date:
2015-04-14 19:34:12 UTC
From:
To:
Am 14.04.2015 um 21:20 schrieb Simon Richter:

I don't see systemd involved here. The interaction is between the
display manager/X and plymouth.
#782609#25
Date:
2015-04-14 19:43:04 UTC
From:
To:
Hi,

systemd needs to get involved in the solution, I think, as this is an
effect of starting services on demand, and is a regression compared to
sysvinit.

If you think otherwise, feel free to drop systemd from the list of
packages this bug applies to, but I doubt this can be solved
satisfactorily without support from the systemd community.

   Simon
#782609#30
Date:
2015-04-14 19:55:03 UTC
From:
To:
Am 14.04.2015 um 21:43 schrieb Simon Richter:

Dropping systemd from the list of affected packages.
You should probably also mention the display manager you are using
#782609#37
Date:
2015-04-14 19:56:28 UTC
From:
To:
Hi,
display manager is gdm.

   Simon
#782609#42
Date:
2015-04-15 10:37:07 UTC
From:
To:
Hi Simon,

so you choose to install a default desktop system and got this? (or did you
first install a non-graphical system and then apt-got into gnome?)


cheers,
	Holger
#782609#47
Date:
2015-04-15 15:01:13 UTC
From:
To:
On Wed, 15 Apr 2015 12:37:07 +0200 Holger Levsen <holger@layer-acht.org> wrote:

Please also mention the version of plymouth and gdm3 you are using,
which theme you selected etc.
#782609#52
Date:
2015-04-16 14:28:01 UTC
From:
To:
Hi Holger,

The latter. These laptops come locked down from the factory with a
BIOS password, so I installed the system by putting the harddisk into
a different machine and running debootstrap, fixing the bootloader and
moving the disk back, then running aptitude install gnome.

Does this make much of a difference for the installation?

My feeling on this bug is that this is more of a general problem
rather than an individual package's fault.

   Simon
#782609#57
Date:
2015-04-16 14:30:13 UTC
From:
To:
Hi,

Everything is from jessie, so that would be plymouth 0.9.0-9, and gdm3
3.14.1-7. Plymouth theme is "lines".

   Simon
#782609#62
Date:
2015-04-17 14:34:28 UTC
From:
To:
Hi,

I cannot reproduce this on a freshly installed jessie (using d-i rc2),
choosing the gnome desktop during installation, neither with or without
plymouth installed.

00:02.0 VGA compatible controller: Intel Corporation Mobile 4 Series Chipset
Integrated Graphics Controller (rev 07)


cheers,
	Holger
#782609#67
Date:
2015-04-21 13:39:11 UTC
From:
To:
Hi,

Am 17.04.2015 um 16:34 schrieb Holger Levsen:

I think that is mainly a matter of overall system performance. As said,
the system in question is a two years old netbook that simply takes that
time to start, but because the services required by X are not started
until required, this means that this part of the boot sequence is moved
to after plymouth has been stopped.

Since plymouth simply leaves the last animation frame on the screen in
this case (so we have a seamless transition to X), it appears as if the
animation hangs.

The basic assumption "starting the X server takes less than a second"
simply is not true anymore when we have socket-activated services, and
required services take their time starting up, so on slower systems we
might want to somehow poke systemd into starting the services required
by X before the handover, or make the handover explicit (e.g. with a
callback from X).

I'm aware this is a nontrivial problem, and a lot of effort for what is
not a hard technical problem, but a user experience issue.

   Simon
#782609#72
Date:
2015-04-21 13:41:35 UTC
From:
To:
the system I tested on is a 6 year old thinkpad x200s, though it had an
(rather old) ssd...