Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Timo Juhani Lindfors]

Timo Jyrinki  writes:
> 2009/6/9 Luca Capello :
>> Nothing is ready, because I need to tag the upstream version as well.
>
> Would it be too early to "deprecate" xserver-glamo in favor of
> xserver-xorg-video-glamo by adding also
> CONFIG_MFD_GLAMO_FB_XGLAMO_WORKAROUND=n to Debian's config.gta02?

Why change the config? You can disable this workaround at runtime
too. Just put that to init.d/xdm?

-Timo

___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Timo Jyrinki]

2009/6/11 Timo Juhani Lindfors :
>> CONFIG_MFD_GLAMO_FB_XGLAMO_WORKAROUND=n to Debian's config.gta02?
>
> Why change the config? You can disable this workaround at runtime
> too. Just put that to init.d/xdm?

Well if it's wrong to do in the kernel anyway it'd be more correct to
disable it in the kernel. But please document that runtime thing
somewhere in the wiki while you are it (I'm unaware of it, I'm just
running my self-compiled kernel with that workaround disabled).

-Timo

___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Luca Capello]

Hi Timo!
On Thu, 11 Jun 2009 18:09:57 +0200, Timo Jyrinki wrote:
> 2009/6/9 Luca Capello :
>> Nothing is ready, because I need to tag the upstream version as well.
>
> Would it be too early to "deprecate" xserver-glamo in favor of
> xserver-xorg-video-glamo by adding also
> CONFIG_MFD_GLAMO_FB_XGLAMO_WORKAROUND=n to Debian's config.gta02?

The full description reads:

  config MFD_GLAMO_FB_XGLAMO_WORKAROUND
bool "Smedia Glamo 336x/337x Xglamo rotation workaround"
depends on MFD_GLAMO_FB
help
  This is a workaround for a Xglamo bug. This should be fixed
  in Xglamo and not in kernel space.

  If unsure, say N.

Can this be related to ?

> Generally I have at least no more problems with a "real" X.org + glamo
> DDX driver than with xserver-glamo,

1) What does DDX mean here?  Specifically, which branch (master,
   dri-aware, exa-via-dri) is the Glamo driver you use built from?

2) Which virtual keyboard do you use?  xserver-xorg-video-glamo and
   matchbox-keyboard do not play well together:

 
http://lists.linuxtogo.org/pipermail/smartphones-userland/2009-May/001483.html

   If the bug is present on any Openmoko image (including the fso ones),
   then it is a matchbox-keyboard bug: since there has not been any new
   SVN commit for matchbox-keyboard, I guess that we should track it
   down (is XTest support the real problem?) and fix it.

Thx, bye,
Gismo / Luca


pgpnKVTTUOrKZ.pgp
Description: PGP signature
___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Luca Capello]

Hi there!

On Fri, 12 Jun 2009 16:46:25 +0200, Luca Capello wrote:
> On Thu, 11 Jun 2009 18:09:57 +0200, Timo Jyrinki wrote:
>> 2009/6/9 Luca Capello :
>>> Nothing is ready, because I need to tag the upstream version as well.
>>
>> Would it be too early to "deprecate" xserver-glamo in favor of
>> xserver-xorg-video-glamo by adding also
>> CONFIG_MFD_GLAMO_FB_XGLAMO_WORKAROUND=n to Debian's config.gta02?
>
> The full description reads:
>
>   config MFD_GLAMO_FB_XGLAMO_WORKAROUND
> bool "Smedia Glamo 336x/337x Xglamo rotation workaround"

I have now found the upstream bug:

  http://docs.openmoko.org/trac/ticket/2255

Until Xglamo will be fixed (I guess never...), unfortunately the
workaround should stay there as per upstream reason:

  http://docs.openmoko.org/trac/ticket/2255#comment:13

However, I agree that it should be off by default, since Xglamo is dead
upstream (IMHO all the efforts should concentrate on the X.org version).

> Can this be related to ?

On the upstream bug above you can read about how to disable the
workaround:

  echo 0 
>/sys/class/i2c-adapter/i2c-0/0-0073/pcf50633-regltr.9/glamo3362.0/glamo-fb.0/xglamo_hack

While testing it I found that the Debian kernel 2.6.29 does not have it,
since I have not merged upstream GTA02 config yet.  However, nothing
change WRT the `xrandr --rotate` bug, even with the Openmoko kernel :-(

Thx, bye,
Gismo / Luca


pgpERdu1eoTtx.pgp
Description: PGP signature
___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Timo Jyrinki]

2009/6/12 Luca Capello :
>          This is a workaround for a Xglamo bug. This should be fixed
>          in Xglamo and not in kernel space.
>
>          If unsure, say N.
>
> Can this be related to ?

Not directly. I heard from Lindi that having that workaround not
disabled causes crash in rotating eventually, but not immediately. I
was able to xrandr -o left xrandr -o normal etc. many times even when
I uses kernel with that workaround enabled.

So I don't really know what's going on with that bug's filer.

> 1) What does DDX mean here?  Specifically, which branch (master,
>   dri-aware, exa-via-dri) is the Glamo driver you use built from?

http://packages.qa.debian.org/x/xf86-video-glamo.html , ie. master but
a bit old, and self-compiled to patch out the spurious EXA/16bit
warnings that fill in Xorg.0.log:

--- xf86-video-glamo-0.0.0+20090224.git703acea1.orig/src/glamo-draw.h
+++ xf86-video-glamo-0.0.0+20090224.git703acea1/src/glamo-draw.h
@@ -28,7 +28,7 @@

 void GLAMOWaitIdle(GlamoPtr *pGlamo);

-#define GLAMO_TRACE_FALL 1
+#define GLAMO_TRACE_FALL 0
 #define GLAMO_TRACE_DRAW 1

 #if GLAMO_TRACE_FALL

> 2) Which virtual keyboard do you use?  xserver-xorg-video-glamo and
>   matchbox-keyboard do not play well together:
>
>     
> http://lists.linuxtogo.org/pipermail/smartphones-userland/2009-May/001483.html

Using matchbox-keyboard but it works for me. I think this workaround
was spread around at some point:

/etc/default/nodm:
NODM_X_OPTIONS='vt3 -nolisten tcp -dpi 96'

And it was probably what fixed the problem. Using Sans 12 font at the
moment for GTK. Obligatory screenshot:
http://users.tkk.fi/~tajyrink/moko/debian4.png

-Timo

___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Timo Juhani Lindfors]

Timo Jyrinki  writes:
> Not directly. I heard from Lindi that having that workaround not
> disabled causes crash in rotating eventually, but not immediately. I
> was able to xrandr -o left xrandr -o normal etc. many times even when
> I uses kernel with that workaround enabled.

Please see "xf86-video-glamo/703acea13: xrandr -o 1; xrandr -o 3
causes distortion" -- http://docs.openmoko.org/trac/ticket/2255

I just added

echo 0 > 
/sys/class/i2c-adapter/i2c-0/0-0073/pcf50633-regltr.9/glamo3362.0/glamo-fb.0/xglamo_hack

to /etc/rc.local to be able to use Xorg with the defconfig_gta02 (I
prefer not to change the default config to make bug reports easier to
reproduce).


___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Luca Capello]

Hi Timo!

On Sun, 14 Jun 2009 10:50:58 +0200, Timo Jyrinki wrote:
> 2009/6/12 Luca Capello :
>> Can this be related to ?
[...]
> So I don't really know what's going on with that bug's filer.

I filed that bug ;-)

According to the xrandr manpage:

RandR version 1.2 options
[...]
Per-output options
[...] [--rotate orientation] [...]

RandR version 1.0 and version 1.1 options
[-o orientation] [...]

Since xf86-video-glamo claims to be RandR 1.2 compatible, the --rotate
option must work as well, which is not the case.

Frankly speaking, I am quite surprised that we get this bug on Debian
only and also that no one seems to care.  The best thing would be to
test this behavior on an Openmoko image, but I do not know which one
already use xf86-video-glamo instead of Xglamo.

>> 1) What does DDX mean here?  Specifically, which branch (master,
>>   dri-aware, exa-via-dri) is the Glamo driver you use built from?
>
> http://packages.qa.debian.org/x/xf86-video-glamo.html , ie. master but
> a bit old,

I have not packaged a newer version yet because of a known regression:

  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=520955#58
  https://docs.openmoko.org/trac/ticket/2281

However, I cannot find upstream Git commit 0f03b39435910, which causes
the problem above.

> and self-compiled to patch out the spurious EXA/16bit warnings that
> fill in Xorg.0.log:
>
> --- xf86-video-glamo-0.0.0+20090224.git703acea1.orig/src/glamo-draw.h
> +++ xf86-video-glamo-0.0.0+20090224.git703acea1/src/glamo-draw.h
> @@ -28,7 +28,7 @@
>
>  void GLAMOWaitIdle(GlamoPtr *pGlamo);
>
> -#define GLAMO_TRACE_FALL 1
> +#define GLAMO_TRACE_FALL 0
>  #define GLAMO_TRACE_DRAW 1
>
>  #if GLAMO_TRACE_FALL

This was changed by upstream as well:

  
http://git.openmoko.org/?p=xf86-video-glamo.git;a=commitdiff;h=22997a5a5d4f53872d3a3073611c0cd7f1bcc263

I can simply include upstream fix or package a newer checkout, but I
would like to not cause any regression.

>> 2) Which virtual keyboard do you use?  xserver-xorg-video-glamo and
>>   matchbox-keyboard do not play well together:
>>
>>     
>> http://lists.linuxtogo.org/pipermail/smartphones-userland/2009-May/001483.html
>
> Using matchbox-keyboard but it works for me.

The problem I was referring to was in the end of the post linked above:
=
Other minor issues:
 - the manpage is the same as that of the framebuffer server
 - for some reason, now matchbox-keyboard has some lag after hitting a
   key, and it somehow goes into autorepeat mode. Alt+D shows the
   desktop then hides it then shows it again and so on, Alt+X opens
   zillions of xterms... this could however be caused by something
   completely unrelated.
=

I experience the latter which renders matchbox-keyboard useless, since
as soon as you try to start an xterm with C-A-x the complete system
becomes unusable.  I am not sure if this is related to XTest support in
xf86-video-glamo or not:

  http://bugs.debian.org/501575

> I think this workaround was spread around at some point:
>
> /etc/default/nodm:
> NODM_X_OPTIONS='vt3 -nolisten tcp -dpi 96'
>
> And it was probably what fixed the problem. Using Sans 12 font at the
> moment for GTK. Obligatory screenshot:
> http://users.tkk.fi/~tajyrink/moko/debian4.png

The "official" workaround is described in the xserver-org-video-glamo
README.Debian, as per

  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=500626#84

Other Matchbox components, -panel and -window-manager, suffer of:

  http://bugs.debian.org/529321
  http://bugs.debian.org/531824

Thx, bye,
Gismo / Luca


pgpnDxLI5mtiw.pgp
Description: PGP signature
___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Luca Capello]

Hi Timo!

On Sun, 14 Jun 2009 12:04:33 +0200, Timo Juhani Lindfors wrote:
> Timo Jyrinki  writes:
>> Not directly. I heard from Lindi that having that workaround not
>> disabled causes crash in rotating eventually, but not immediately. I
>> was able to xrandr -o left xrandr -o normal etc. many times even when
>> I uses kernel with that workaround enabled.
>
> Please see "xf86-video-glamo/703acea13: xrandr -o 1; xrandr -o 3
> causes distortion" -- http://docs.openmoko.org/trac/ticket/2255

As I wrote previously, I think the Xglamo workaround should be off by
default, this is being discussed at:

  http://thread.gmane.org/gmane.comp.handhelds.openmoko.kernel/9960

> I just added
>
> echo 0 > 
> /sys/class/i2c-adapter/i2c-0/0-0073/pcf50633-regltr.9/glamo3362.0/glamo-fb.0/xglamo_hack
>
> to /etc/rc.local to be able to use Xorg with the defconfig_gta02

The "correct" option to modify sysfs files would be to install
sysfsutils and then edit /etc/sysfs.conf ;-)

Thx, bye,
Gismo / Luca


pgp0Z8c7IRUs7.pgp
Description: PGP signature
___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Timo Jyrinki]

2009/6/14 Luca Capello :
>        [...] [--rotate orientation] [...]
>        [-o orientation] [...]
>
> Since xf86-video-glamo claims to be RandR 1.2 compatible, the --rotate
> option must work as well, which is not the case.

Ah, no wonder I was puzzled about "rotations not working". Yes indeed
per-output setting should work as well.

> test this behavior on an Openmoko image, but I do not know which one
> already use xf86-video-glamo instead of Xglamo.

I think the answer is "none", because of proper X.org packaging
lacking in OpenEmbedded, but not sure.

> However, I cannot find upstream Git commit 0f03b39435910, which causes
> the problem above.

http://git.openmoko.org/?p=xf86-video-glamo.git;a=commit;h=0f03b39435910933a34f5b5dccbc57fb55a2b6f1

There has been 7 commits after that, so maybe people should also test
the newest version.. On the other hand, I'd already be interested in
Thomas's DRI/DRM/EXA stack of
http://git.openmoko.org/?p=kernel.git;a=shortlog;h=refs/heads/drm-tracking
+ http://git.bitwiz.org.uk/?p=libdrm.git;a=shortlog;h=refs/heads/glamo
+ 
http://git.openmoko.org/?p=xf86-video-glamo.git;a=shortlog;h=refs/heads/exa-via-dri
because that's a) cool b) something that with addition of KMS could
offer much better user experience in the long run... if there is a
possibility to get rid of both crashes under load and busy loops that
stall the CPU, Glamo could actually become a sort of accelerator it's
currently not :)

>  http://git.openmoko.org/?p=xf86-video-glamo.git;a=commitdiff;h=22997a5a5d4f53872d3a3073611c0cd7f1bcc263
>
> I can simply include upstream fix or package a newer checkout, but I
> would like to not cause any regression.

Yes, it'd be beneficial for -2 if there is uncertainty about whether
to upgrade to later git version.

>  - for some reason, now matchbox-keyboard has some lag after hitting a
>   key, and it somehow goes into autorepeat mode. Alt+D shows the
>   desktop then hides it then shows it again and so on, Alt+X opens
>   zillions of xterms... this could however be caused by something
>   completely unrelated.

Yeah this is also a true problem. Too bad there are no fixes but only
workarounds to these (at least) two problems, ie. font size
(workaround -dpi 96) and keys repeating (workaround disabling key
repeat which I'm using).

-Timo

___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Luca Capello]

Hi Timo!

On Mon, 15 Jun 2009 13:26:23 +0200, Timo Jyrinki wrote:
> 2009/6/14 Luca Capello :
>> However, I cannot find upstream Git commit 0f03b39435910, which causes
>> the problem above.
>
> http://git.openmoko.org/?p=xf86-video-glamo.git;a=commit;h=0f03b39435910933a34f5b5dccbc57fb55a2b6f1

Thanks, stupid me :-)

> There has been 7 commits after that, so maybe people should also test
> the newest version..

This is my idea: testing each new commit and find out if the xrandr bug
has been fixed and if new bugs appear.

> On the other hand, I'd already be interested in
> Thomas's DRI/DRM/EXA stack of
[...]
> because that's a) cool b) something that with addition of KMS could
> offer much better user experience in the long run... if there is a
> possibility to get rid of both crashes under load and busy loops that
> stall the CPU, Glamo could actually become a sort of accelerator it's
> currently not :)

The major problem here is that we need too many extra stuff: a new
kernel, a new libdrm and then a new xf86-video-glamo.  Moreover, since
libdrm is already in Debian main, we should coordinate any new package
upload with its maintainer.

Finally, is someone actually working on Glamo?  Its life is directly
linked with GTA02, since, as it was planned for GTA03, the new
gta02-core community project has removed it.

>>  http://git.openmoko.org/?p=xf86-video-glamo.git;a=commitdiff;h=22997a5a5d4f53872d3a3073611c0cd7f1bcc263
>>
>> I can simply include upstream fix or package a newer checkout, but I
>> would like to not cause any regression.
>
> Yes, it'd be beneficial for -2 if there is uncertainty about whether
> to upgrade to later git version.

I will firstly perform the tests I described above and then evaluate
what to do: in any case, the next upload will include that fix.

Thx, bye,
Gismo / Luca


pgphHLEbploNJ.pgp
Description: PGP signature
___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland

Re: Deprecating xserver-glamo (was: Recording from headset when pressing headset button

[Timo Jyrinki]

2009/6/24 Luca Capello :
> The major problem here is that we need too many extra stuff: a new
> kernel, a new libdrm and then a new xf86-video-glamo.  Moreover, since
> libdrm is already in Debian main, we should coordinate any new package
> upload with its maintainer.

If it gets ready this is all doable. But at the moment that stuff is
purely for user testing if something works or not.

> Finally, is someone actually working on Glamo?  Its life is directly
> linked with GTA02, since, as it was planned for GTA03, the new
> gta02-core community project has removed it.

Regarding working on development, Lars-Peter at least has again done a
few ioctl-related commits. It's been a few weeks since we've heard
from Thomas (AFAIK), but he seemed enthusiastic about experimenting
with KMS & OpenGL.

Whatever the case, it is probable FreeRunner will still be the only
truly Free phone for the next two years or so at least. Gta02-core
might not ever produce an actual hw product, and the other Linux
phones are not as hackable. Currently the glamo operation is not rock
stable, and is loading CPU too much - the major usage improvement for
all of us would be to get KMS into stable state, in which case glamo
can be guaranteed the regular calls it requires for functioning
properly (user-space cannot guarantee this). At the same time KMS
could be, in my very limited understanding, the key to not stalling
CPU on gfx operations by not busy-looping anymore.

Regarding Debian, nothing needs to be done at the moment regarding
libdrm, kernel and mesa. But if the time comes they are relatively
stable, I'll be happy to help in packaging the glamo bits.

-Timo

___
Smartphones-userland mailing list
Smartphones-userland@linuxtogo.org
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/smartphones-userland