changeset 16163: drivers-dvbsky

[Rudi]

Hi Tom,

last night you added the mentioned package. I'm not happy with this :-(.
This is because it's not just a driver for the "dvbsky" hardware but a
replacement for the whole v4l driver subtree ! It overlays (and possibly
even downlevels) drivers that are already part of the kernel.

Therefore I would vote for a different approach:

On the web site there is this file:

http://www.dvbsky.eu/download/linux-3.6.6-dvbsky.patch.tar.gz

The patch applies on 3.6 and 3.5 kernels (RPi and CuBox) and most likely also
on the kernels used on PC targets. I guess this does less harm than replacing
all the media drivers with a tarball of questionable source. The down side is,
that the patch fails on Panda (kernel 3.4)...



--

Ruediger "Rudi" Ihle

[Thomas Genty]

Le 19/05/2013 09:24, Rudi a �crit :
> Hi Tom,

Hello Rudi

>
> last night you added the mentioned package. I'm not happy with this :-(.
> This is because it's not just a driver for the "dvbsky" hardware but a
> replacement for the whole v4l driver subtree ! It overlays (and possibly
> even downlevels) drivers that are already part of the kernel.

true: note is was added as a module so if someone wants to install this
one, it's because he has such device

>
> Therefore I would vote for a different approach:
>
> On the web site there is this file:
>
> http://www.dvbsky.eu/download/linux-3.6.6-dvbsky.patch.tar.gz
>
> The patch applies on 3.6 and 3.5 kernels (RPi and CuBox) and most
> likely also
> on the kernels used on PC targets.

humm ,not on i386 I guess, not the same kernel version

> I guess this does less harm than replacing
> all the media drivers with a tarball of questionable source. The down
> side is,
> that the patch fails on Panda (kernel 3.4)...

Ok, I prefer this approach too
(didn't think this patch applied on 3.5)
we can add this patch on defconfig/platform, agree ?
then it won't break panda
>
>
>
Regards

Tom

[Rudi]

Hello,

> true: note is was added as a module so if someone wants to install this one, it's because he has such device

I know. I read the forum from time to time,,,

> humm ,not on i386 I guess, not the same kernel version

I always thought that the PC stuff is more up to date than our SoCs... Maybe I'm wrong ;-).

> (didn't think this patch applied on 3.5)

I was just curious and it succeeded with some lines offset...

> we can add this patch on defconfig/platform, agree ?

Yes.


--

Ruediger "Rudi" Ihle

[Thomas Genty]

Le 19/05/2013 10:13, Rudi a �crit :

> Hello,
>
>> true: note is was added as a module so if someone wants to install
>> this one, it's because he has such device
>
> I know. I read the forum from time to time,,,
>

:-)

>
>> humm ,not on i386 I guess, not the same kernel version
>
> I always thought that the PC stuff is more up to date than our SoCs...
> Maybe I'm wrong ;-).
>

need to test a more recent patch for kernel 3.8.10

>
> > (didn't think this patch applied on 3.5)
>
> I was just curious and it succeeded with some lines offset...
>

seems to build fine

>
>> we can add this patch on defconfig/platform, agree ?
>
> Yes.
>
>

Added for Cubox.

Cheers

Tom

[Rudi]

On 19.05.2013 10:43, Thomas Genty wrote:

> seems to build fine
> ...
> Added for Cubox.

Let's see if it works for the guy who requested it. If so, I guess it
won't hurt to add it for RaspberryPi as well.

BTW, a while back we talked about upgrading the CuBox kernel. AFAIR,
Rabeeh has a 3.6.9 or so. However, at the moment I'm not so sure if
we should try that. It looks like the 3.8 / 3.9 mainline kernels are
making some good progress. Since our 3.5.7 is performing pretty well,
I see no real need to switch. Instead I would prefer to wait for a
3.9+ to become more feature-complete.


--

Ruediger "Rudi" Ihle

[Thomas Genty]

Le 19/05/2013 11:30, Rudi a �crit :

> On 19.05.2013 10:43, Thomas Genty wrote:
>
>> seems to build fine
> > ...
>> Added for Cubox.
>
> Let's see if it works for the guy who requested it. If so, I guess it
> won't hurt to add it for RaspberryPi as well.

;-)
Waiting for the bot to start a new cycle.

>
> BTW, a while back we talked about upgrading the CuBox kernel. AFAIR,
> Rabeeh has a 3.6.9 or so. However, at the moment I'm not so sure if
> we should try that. It looks like the 3.8 / 3.9 mainline kernels are
> making some good progress. Since our 3.5.7 is performing pretty well,
> I see no real need to switch. Instead I would prefer to wait for a
> 3.9+ to become more feature-complete.

Ok , so wait and see.


Tom

[Rudi]

Hi Tom,

do we have a tool or script to update the other defconfigs automatically after adding new packages or platforms ?

BTW, with the "base" config the A10 boots nicely to a command prompt. Next would be XBMC. However, I'm afraid that I won't find the time to try this within the next two weeks :-(

[Thomas Genty]

Le 25/05/2013 09:31, Rudi a �crit :
> Hi Tom,
Hello Rudi,

>
> do we have a tool or script to update the other defconfigs automatically after adding new packages or platforms ?

yes : refreshcfg
./scripts/refreshcfg

>
> BTW, with the "base" config the A10 boots nicely to a command prompt. Next would be XBMC. However, I'm afraid that I won't find the time to try this within the next two weeks :-(
>
>

no hurry ;-)

Regards,

Tom

[Rudi]

Hi Tom,

> yes : refreshcfg
> ./scripts/refreshcfg

Ahh. That helps a lot ;-).

Next question: Do we have a tool that list the packages that reference a
given one ?

>> BTW, with the "base" config the A10 boots nicely to a command prompt.
>> Next would be XBMC. However, I'm afraid that I won't find the time to
>> try this within the next two weeks :-(
>>
>>
> no hurry ;-)

We'll see what happens. It looks like there is an XBMC version out there
for the A10. However - as ususal - it seems to have some problems...


BTW, somewhere on the web I read, that using LZO compression for the
kernel would speed up boot times a bit. Even without the massive
improvements to the LZO code that have been comitted to kernel 3.9
recently. I'm wondering, if we should give that a try...



--

Ruediger "Rudi" Ihle

[Thomas Genty]

Le 26/05/2013 12:28, Rudi a �crit :

> Hi Tom,
>
>> yes : refreshcfg
>> ./scripts/refreshcfg
> Ahh. That helps a lot ;-).

;-)

>
> Next question: Do we have a tool that list the packages that reference a
> given one ?

Not sure to understand what you want exactly ...
we have opkg that give installed package ...

>
>>> BTW, with the "base" config the A10 boots nicely to a command prompt.
>>> Next would be XBMC. However, I'm afraid that I won't find the time to
>>> try this within the next two weeks :-(
>>>
>>>
>> no hurry ;-)
> We'll see what happens. It looks like there is an XBMC version out there
> for the A10. However - as ususal - it seems to have some problems...

heh, nobody is perfect

>
>
> BTW, somewhere on the web I read, that using LZO compression for the
> kernel would speed up boot times a bit. Even without the massive
> improvements to the LZO code that have been comitted to kernel 3.9
> recently. I'm wondering, if we should give that a try...

why not ....
I'll try this this afternoon if I have enough spare time
but boot times takes me about 3s on my p4... and 10s from bios to
syslinux so ... the problem (for me) is not about kernel.

Have a nice week-end

Regards

Tom

[Rudi]

>> Next question: Do we have a tool that list the packages that reference a
>> given one ?
> Not sure to understand what you want exactly ...
> we have opkg that give installed package ...

What I mean is the following: In our meta information we have specified
package "A" as a dependency (no matter if build or runtime) for packages
"B" and "C". Now I would like to call something like:

"./scripts/list_ref A"

which would geve me the answer "B" and "C".


--

Ruediger "Rudi" Ihle

[Thomas Genty]

Le 26/05/2013 14:35, Rudi a �crit :

>>> Next question: Do we have a tool that list the packages that reference a
>>> given one ?
>> Not sure to understand what you want exactly ...
>> we have opkg that give installed package ...
> What I mean is the following: In our meta information we have specified
> package "A" as a dependency (no matter if build or runtime) for packages
> "B" and "C". Now I would like to call something like:
>
> "./scripts/list_ref A"

no, no such script

>
> which would geve me the answer "B" and "C".
>
>

something like that (for mesa]
grep mesa packages/*/meta | grep "PKG_DEPENDS\|PKG_BUILD_DEPENDS" | cut
-d/ -f2 ?

Tom

[Mike Valk]

On Sunday, May 26, 2013 12:28:12 PM UTC+2, Rudi wrote:

Hi Tom,

> yes : refreshcfg
> ./scripts/refreshcfg

Ahh. That helps a lot ;-).

Next question: Do we have a tool that list the packages that reference a
given one ?


>> BTW, with the "base" config the A10 boots nicely to a command prompt.
>> Next would be XBMC. However, I'm afraid that I won't find the time to
>> try this within the next two weeks :-(
>>
>>
> no hurry ;-)

We'll see what happens. It looks like there is an XBMC version out there
for the A10. However - as ususal - it seems to have some problems...

I've been following the progress on XBMC for A10 as well. It seems that the linux version of the CedarX library is buggy. The Android one seems to be better. Apparently someone managed to use that one with libHybris
http://forum.xbmc.org/showthread.php?tid=126995&pid=1419678#pid1419678

But other than that not much progress seems to be on route.

There are some activities going for reverse engineering
https://groups.google.com/forum/#!forum/linux-sunxi

Also the lima drivers are on the brink of being usable.

Good luck anyway.
 

[Rudi]

On 29.05.2013 10:57, Mike Valk wrote:

> I've been following the progress on XBMC for A10 as well. It seems that the
> linux version of the CedarX library is buggy. The Android one seems to be better.
> Apparently someone managed to use that one with libHybris
> http://forum.xbmc.org/showthread.php?tid=126995&pid=1419678#pid1419678

Thanks for the info.

> There are some activities going for reverse engineering
> https://groups.google.com/forum/#!forum/linux-sunxi

I guess that will take some time...

> Also the lima drivers are on the brink of being usable.

Are the original MALI400 drivers buggy as well ?



--

Ruediger "Rudi" Ihle

[Mike Valk]

Have not read as such. But changing kernels can only be done when ARM releases a new driver. And the Android driver takes precedence over linux/x11,fb for ARM.

The MALI driver has some quirks. Hence the effort from Luc to build an open driver.

Also Luc claims to have a gain in speed and efficiency over the binary driver. The binary driver probably is the collection of 3rd party code, hence the inability to release code from ARM.
 

--

Ruediger "Rudi" Ihle