[pkg-fso-maint] Bug#860463: nodm: Missing handling of plymouth (#782456)

[Simon McVittie]

On Mon, 17 Apr 2017 at 12:05:41 +0200, Niels Thykier wrote:
> For init scripts that can be done by the following shell snippet (just
> before starting the DM):
> 
> """
> if [ -x /bin/plymouth ] ; then
>    /bin/plymouth quit
> fi
> """

Should this say "/bin/plymouth quit --wait"?
https://bugs.launchpad.net/ubuntu/+source/nodm/+bug/714492 suggests that
maybe it should, but you seem fairly certain that it shouldn't, and I'm
not sure I understand why?

> For systemd, it should be a question of adding the following under the
> "[Unit]" section:
> 
> """
> After=plymouth-quit.service
> """

Similarly, should this additionally be After plymouth-quit-wait.service?
I'm not sure that I see how it helps to guarantee that we already asked
plymouth to relinquish control over the screen, if we have not also
waited for it to do so?

On the other hand,
https://src.fedoraproject.org/cgit/rpms/nodm.git/tree/nodm.service
(installed by the distro where plymouth originated) does agree with what
you're suggesting here.

The reason I submitted the pile of patches that led to me being added as
a nodm co-maintainer (without having asked for that position) is that my
employer is using nodm for the SDK virtual machine image of an embedded
Debian derivative, with a service file similar to what I sent upstream. We
use plymouth in that system, and we have not had problems with booting
that system, leading me to believe that we are not actually seeing the
symptoms of #860463 there; so this clearly doesn't affect all systems,
and is going to be hard to test meaningfully.

Looking at other *DM that do not have bugs opened against them:

gdm, lightdm and lxdm cannot be used as examples because they stop
Plymouth in C code to get more control over it, which is entirely
valid, but not applicable to simpler DMs like nodm that do not do
the same.

sddm doesn't seem to order itself after plymouth-quit.service or
plymouth-quit-wait.service in the systemd code path, although it does
ask plymouth to quit (but does not wait) in the sysvinit/LSB code path.

xdm doesn't seem to do anything plymouth-related either. Is that another
instance of this bug? Fedora does patch it to order itself after
plymouth-quit.service:
https://src.fedoraproject.org/cgit/rpms/xorg-x11-xdm.git/tree/xdm-service.patch

    S