Do we really want shared Xserver script for all machines?

Do we really want shared Xserver script for all machines?

[Martin Jansa]

Hi,

on Openmoko Freerunner om-gta02 Xserver script takes about 3s on every
boot
http://build.shr-project.org/tests/jama/bootchart-2.6.32.9.r1.png

it's started from /etc/init.d/xserver-nodm and with eliminated Xserver
call by passing xinit call directly with right Xorg params I get this
 
http://build.shr-project.org/tests/jama/bootchart-2.6.32.9.r3.png

My question is:

should I commit simplified versions of Xserver script
xserver-kdrive-common/xserver-kdrive-common/Xserver
to
xserver-kdrive-common/xserver-kdrive-common/om-gta02/Xserver
and the same for other devices supported in Xserver (to save 
grep,awk call).

or do it only for om-gta02 (maybe even in 
xserver-kdrive-common/xserver-nodm-init/xserver-nodm
) and those who care about boot time will follow?

Maybe calling fallback_screen_arg only where it's really needed would be
good enough optimization, without the cost of loosing PACKAGE_ARCH =
"all".

Cheers,

-- 
uin:136542059                jid:Martin.Jansa@gmail.com
Jansa Martin                 sip:jamasip@voip.wengo.fr 
JaMa                         

Re: Do we really want shared Xserver script for all machines?

[Koen Kooi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Wouldn't it be better to merge the scripts from xserver-kdrive-common
and xserver-common and then fix all the sed/awk calls in one place?

On 16-04-10 14:39, Martin Jansa wrote:
> Hi,
> 
> on Openmoko Freerunner om-gta02 Xserver script takes about 3s on every
> boot
> http://build.shr-project.org/tests/jama/bootchart-2.6.32.9.r1.png
> 
> it's started from /etc/init.d/xserver-nodm and with eliminated Xserver
> call by passing xinit call directly with right Xorg params I get this
>  
> http://build.shr-project.org/tests/jama/bootchart-2.6.32.9.r3.png
> 
> My question is:
> 
> should I commit simplified versions of Xserver script
> xserver-kdrive-common/xserver-kdrive-common/Xserver
> to
> xserver-kdrive-common/xserver-kdrive-common/om-gta02/Xserver
> and the same for other devices supported in Xserver (to save 
> grep,awk call).
> 
> or do it only for om-gta02 (maybe even in 
> xserver-kdrive-common/xserver-nodm-init/xserver-nodm
> ) and those who care about boot time will follow?
> 
> Maybe calling fallback_screen_arg only where it's really needed would be
> good enough optimization, without the cost of loosing PACKAGE_ARCH =
> "all".
> 
> Cheers,
> 

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFLyGVUMkyGM64RGpERAiyrAJwJHKahFvUdK8l+aYIgBhb2UMlwxwCgkzEF
sXq4fCHgeNYcSUlIXT/aKp4=
=Ft6E
-----END PGP SIGNATURE-----

Re: Do we really want shared Xserver script for all machines?

[Marcin Juszkiewicz]

Dnia piątek, 16 kwietnia 2010 o 15:25:40 Koen Kooi napisał(a):
>   Wouldn't it be better to merge the scripts from xserver-kdrive-common
> and xserver-common and then fix all the sed/awk calls in one place?

I did most of it long time ago and stored as xserver-common 1.30. Script
got simplified, supports Xorg and KDrive, replaces xserver-kdrive-common,
allows to set variables in /etc/default/xserver instead of patching 
/etc/X11/Xserver file etc.

I did send RFC about it on 3rd August 2009. Merged it into stable/2009 in
October 2009 as we used it on BUG 1.x and 2.0 devices.

I merged my improvements into upstream which gave us 1.31, 1.32 and 1.33
versions (1.30/upstream != 1.30/OE as it was base for my improvements which
landed in next upstream versions).

So far I did not noticed anyone using it. For those which use 
xserver-nodm-init script to boot into root X11 session: you need newer
version from BugLabs repository:

svn://svn.buglabs.net/bug/trunk/com.buglabs.build.oe/meta-bug/packages/xserver-nodm-init

Regards, 
-- 
JID:      hrw@jabber.org
Website:  http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz

Re: Do we really want shared Xserver script for all machines?

[Martin Jansa]

On Fri, Apr 16, 2010 at 03:41:22PM +0200, Marcin Juszkiewicz wrote:
> I did send RFC about it on 3rd August 2009. Merged it into stable/2009 in
> October 2009 as we used it on BUG 1.x and 2.0 devices.
> 
> So far I did not noticed anyone using it. For those which use 

I guess this is because task-x11-utils pulls xserver-kdrive-common
unconditionaly.. I've added variable there for distro to pick provider
easily and after some test I'll send it for review.

> xserver-nodm-init script to boot into root X11 session: you need newer
> version from BugLabs repository:
> 
> svn://svn.buglabs.net/bug/trunk/com.buglabs.build.oe/meta-bug/packages/xserver-nodm-init

In patch I have now, I renamed old version to 1.0 and added version from
buglabs as 2.0 with D_P -1. Not best solution as it allows people to use
old xserver-kdrive-common + this newer nodm-init, but comment above D_P
setting should be enough for maintainers to notice.

Thanks hrw, pity that we overlooked xserver-common for so long..

Regards,

-- 
uin:136542059                jid:Martin.Jansa@gmail.com
Jansa Martin                 sip:jamasip@voip.wengo.fr 
JaMa                         

Re: Do we really want shared Xserver script for all machines?

[Martin Jansa]

On Fri, Apr 16, 2010 at 03:41:22PM +0200, Marcin Juszkiewicz wrote:
> Dnia piątek, 16 kwietnia 2010 o 15:25:40 Koen Kooi napisał(a):
> >   Wouldn't it be better to merge the scripts from xserver-kdrive-common
> > and xserver-common and then fix all the sed/awk calls in one place?

Hi,

it's better with xserver-common, here are bootcharts again from
Freerunner

http://build.shr-project.org/tests/jama/bootchart-2.6.32.11.xserver-common.png

and without udevd/hald combo:
http://build.shr-project.org/tests/jama/bootchart-2.6.32.11.xserver-common.no-udev.png

I'll add /etc/X11/Xsession.d/90xXWindowManager and
/etc/X11/Xsession.d/89xTs_Calibrate from xserver-kdrive-common to
xserver-common (if you don't mind or have better place for them) and
push it (enabled by default for SHR).

Thanks,

-- 
uin:136542059                jid:Martin.Jansa@gmail.com
Jansa Martin                 sip:jamasip@voip.wengo.fr 
JaMa                         

Re: Do we really want shared Xserver script for all machines?

[Sergey Lapin]

On Fri, Apr 16, 2010 at 4:39 PM, Martin Jansa <martin.jansa@gmail.com> wrote:
> Hi,
>
> on Openmoko Freerunner om-gta02 Xserver script takes about 3s on every
> boot
> http://build.shr-project.org/tests/jama/bootchart-2.6.32.9.r1.png
>
> it's started from /etc/init.d/xserver-nodm and with eliminated Xserver
> call by passing xinit call directly with right Xorg params I get this
>
> http://build.shr-project.org/tests/jama/bootchart-2.6.32.9.r3.png

Sorry to bother with off-topic post, but how these boot charts are
built? Really nice thing to have...

[OT] bootchart (was: Do we really want shared Xserver script for all machines?)

[Paul Menzel]

[-- Attachment #1: Type: text/plain, Size: 261 bytes --]

Am Freitag, den 16.04.2010, 17:39 +0400 schrieb Sergey Lapin:

[…]

> Sorry to bother with off-topic post, but how these boot charts are
> built? Really nice thing to have...

Use bootchart [1]?


Thanks,

Paul


[1] http://www.bootchart.org/

[-- Attachment #2: Dies ist ein digital signierter Nachrichtenteil --]
[-- Type: application/pgp-signature, Size: 205 bytes --]

Re: [OT] bootchart (was: Do we really want shared Xserver script for all machines?)

[Martin Jansa]

On Fri, Apr 16, 2010 at 11:00:06PM +0200, Paul Menzel wrote:
> Am Freitag, den 16.04.2010, 17:39 +0400 schrieb Sergey Lapin:
> 
> […]
> 
> > Sorry to bother with off-topic post, but how these boot charts are
> > built? Really nice thing to have...
> 
> Use bootchart [1]?

yes,

and to be OE specific I used
recipes/bootchart-lite/bootchart-lite_svn.bb

and for freerunner slightly changed Qi bootloader to pass it as init.
http://gitorious.org/~jama/angstrom/jama-shr-experimental/commit/03c69b54cf1675a30e2d71d36e4662611982a5c6

then tar and scp resulting files back to computer and run bootchart to
generate graph.

-- 
uin:136542059                jid:Martin.Jansa@gmail.com
Jansa Martin                 sip:jamasip@voip.wengo.fr 
JaMa