On Tue, Mar 13, 2018 at 12:18:47AM +0100, Pavel Machek wrote:
> > >> > >> Fully charged
> > >> > >> battery shows as empty battery with red "x" in Mate desktop. After
> > >> > >> reboot to recent mainline problem goes away. Tooltip says "battery
> > >> > >> waiting to charge".
> > >> > >>
> > >>
> > >> For the MATE bug see:
> > >> https://github.com/mate-desktop/mate-power-manager/pull/251
> > >>
> > >> The problem is in the new composite battery system in UPower, it
> > >> "merges" two external batteries into a single one for reporting. That
> > >> battery does not understand "Not Charging" inside the MATE Power
> > >> Manager. The above patch fixes that in MATE.
> > >
> > > Ok, so problem is 91eea70e5e5ce12eb1c7cd922e561fab43e201bd.
> > >
> > > And now, I see that MATE should be improved.
> > >
> > > But what I also see is that kernel is changing its interface in a way
> > > that (while making sense) breaks existing application. And we don't
> > > break existing applications.
> > >
> > > [Now... maybe if the fix for the application is available for long
> > > long time and noone is really looking and everyone already has new
> > > application version, maybe ... But pull request for MATE is 9 days
> > > old, and not even merged. So there is no way this is suitable for
> > > 4.17.]
It's not a bug in the kernel, it's a very subtle cosmetic bug in MATE.
If you can, please test that MATE patch and report back to Github
if it works, as I wrote that patch and they are waiting for someone with
a ThinkPad to test the patch. If it works, it will get merged and
everyone will be happy.
This is far from "breaking applications", it's just a cosmetic bug with
a single icon in a single application that has a simple fix.
The upstream underlying UPower system reports it properly, as
UP_DEVICE_STATE_PENDING_CHARGE, it's just that the MATE/GNOME developers did
not handle that scenario due to poor testing and use case analysis, even
when that state is in the UpDeviceState enumeration.
UPower correctly returns UP_DEVICE_STATE_PENDING_CHARGE, its just that
the MATE Power Manager discards that and assumes
> > Right.
> > This basically means that we will have to defer the battery changes
> > until user space is ready and not just upstream, but also in distro
> > packages.
> Yes please.
No. See below.
> Note that only 91eea70e5e5ce12eb1c7cd922e561fab43e201bd should break
> anything, the others should be safe AFAICT?
Again, nothing is broken here, the kernel is *currently* broken as it
shows that the battery state is "Unknown" and MATE assumes a broken
kernel and returns a simple charging icon.
> (english) http://www.livejournal.com/~pavelmachek
> (cesky, pictures)