trying 41001qq

[James Cameron]

On Sun, Feb 15, 2015 at 01:21:41PM +0000, Yioryos Asprobounitis wrote:
> It is nice to see a new build for all the XOs after some time now.
> Tried 41001qq at the ends of the spectrum my G1G1 XO-1 and a touch
> XO-4

41001qq4 is based on my work.

> Both worked and I even got them to collaborate on measure activity
> On the XO-4 things are really nice and with the SMP kernel things
> are moving reasonably fast.  Sugar runs just fine and activities do
> not misbehave.

Thanks.

> Mate desktop is responsive albeit touch unfriendly. Not sure if
> anything could be done about it as I do not think they have a
> touch-friendly theme or layout.

Won't fix.

> The default 10 point font is too big and 7 or 8 would be better.

Won't fix.  This size was requested by customer.  It seems fine to me.

> Also the switch user option does nothing and should be probably
> hidden, as well as Xterm that is useless on Mate. But these are
> minor tittle things to wary about.

Won't fix.

> On the XO-1, I would say simple do not build for this hardware.

That's a bit harsh.  Samuel has already done initial diagnosis of this
problem, and it was in his known problems list.

> It took me 3 tries and a long unexpected phonecall to actually boot
> in the XO-1, as in the first 2 tries after a minute or 2 with a
> black (power management kicks in) or white screen, I was rebooting
> and checking that everything was OK.
> It is just takes a very looooong time to get going. This is also
> true when starting activities as with most the "fail to start"
> timeout kicks in before the activity has the chance to start. If you
> ignore it, the activity eventually starts.

Yes, that's the effect of the problem Samuel found.
See https://dev.laptop.org/ticket/12860

> Adding a bit of swap did not improve things.

Yes, swap should have no effect on this problem.

> Looking at CPU usage it would appear that systemd-journal consumes
> anything from 35 to 55% of the CPU with just 'top' running and the
> XO-1 is constantly on 3.0+ load average.
> I do not know if systemd-journal can be turned off but would be
> desirable for XO-1 builds.

systemd-journal is handling messages being emitted by wpa_supplicant.

Turning off systemd-journal would (a) lose the messages and any chance
of diagnosing the problem, and (b) make wpa_supplicant spin faster.

It would be better to fix the underlying problem with wpa_supplicant
or Sugar.

> The other heavy CPU user (that I do not remember being the case in
> older builds) is Network Manager/wpa_suplicant with ~20% each.

> Anyway, the build "works" on the XO-1 but with *a lot* of patience.
> So unless the new builds have this "killer/must have" feature, I
> wouldn't bother with XO-1 Fedora/Sugar builds.

It's not what I call a feature.  It's a bug.

My guess is that it's a Sugar or Network Manager bug.

Have you been able to investigate the bug any further?  I'd like to
see community assistance with this one, because I can't justify such
work.  If you can prove it happens on XO-4, then I can certainly look
into it, but I've not noticed a problem there.

-- 
James Cameron
http://quozl.linux.org.au/