Status of ARM machines in OE

Status of ARM machines in OE

[Marcin Juszkiewicz]

During weekend I build 'task-base' for all ARM (armv4t and better) 
machines and Ångström distribution. The list included:

a1200 a780 amsdelta compulab-pxa270 devkitidp-pxa255 ep93xx h1910 h1940 
h2200 h3900 h4000 h5000 h6300 htcblueangel htcuniversal hx2000 hx4700 
ks8695 logicpd-pxa270 magician mainstone mtx-3 mx21ads mx31ads 
navman-icn330 netbook-pro nokia770 omap1510inn omap1610h2 omap1710h3 
omap2420h4 omap5912osk palmld palmtx rokr-e2 rx3000 smdk2440 xxs1500

This week ARM 'broken' machines are:

a1200 - lack of kernel config - bug #1766
omap1710h3 - lack of kernel config - bug #1736
omap2420h4 - lack of kernel config - bug #1744
xxs1500 - unfetchable kernel - bug #1737

=============================================================================

compulab-pxa270 - use binary driver for flash so only hardfloat 
distributions
can use this machine:

arm-angstrom-linux-gnueabi-ld: ERROR: drivers/block/cl_flash/CL_FlashDrv uses FPA instructions, whereas drivers/block/cl_flash/fdrv.o does not
arm-angstrom-linux-gnueabi-ld: failed to merge target specific data of file drivers/block/cl_flash/CL_FlashDrv

Full log: 
http://ewi546.ewi.utwente.nl/tinderbox/showlog.pl?machine_id=177&logfile=20070107145519.log

=============================================================================

h1910 - kernel broken:
arch/arm/mach-pxa/built-in.o: In function `h4000_set_led':
dma_needs_bounce.c:(.text+0x1ca0): undefined reference to `ipaq_asic3_set_led'
dma_needs_bounce.c:(.text+0x1cd0): undefined reference to `ipaq_asic3_set_led'
dma_needs_bounce.c:(.text+0x1ce4): undefined reference to `ipaq_asic3_set_led'
dma_needs_bounce.c:(.text+0x1d04): undefined reference to `ipaq_asic3_set_led'
dma_needs_bounce.c:(.text+0x1d18): undefined reference to `ipaq_asic3_set_led'
make: *** [.tmp_vmlinux1] Error 1

Full log: 
http://ewi546.ewi.utwente.nl/tinderbox/showlog.pl?machine_id=177&logfile=20070107150654.log

=============================================================================

htcblueangel - kernel broken is same place as h1910

=============================================================================

rx3000 - kernel broken is same place as h1910

=============================================================================

htcuniversal - kernel broken:
In file included from lib/audit.c:8:
include/asm-generic/audit_dir_write.h:9: error: '__NR_mkdirat' undeclared here (not in a function)
include/asm-generic/audit_dir_write.h:10: error: '__NR_mknodat' undeclared here (not in a function)
include/asm-generic/audit_dir_write.h:11: error: '__NR_unlinkat' undeclared here (not in a function)
include/asm-generic/audit_dir_write.h:12: error: '__NR_renameat' undeclared here (not in a function)
include/asm-generic/audit_dir_write.h:13: error: '__NR_linkat' undeclared here (not in a function)
include/asm-generic/audit_dir_write.h:14: error: '__NR_symlinkat' undeclared here (not in a function)

Full log: 
http://ewi546.ewi.utwente.nl/tinderbox/showlog.pl?machine_id=177&logfile=20070107173621.log

=============================================================================

magician is OK - builds without problem. But as this is HTC Magician then
maybe we should rename it to htcmagician to follow other htc phones
(htcblueangel, htcuniversal) which are in OE?

=============================================================================

mtx-3 is problematic. According to machine config file this is config of
experimental machine which is not released in hardware yet. In that file
it has enabled optimalizations for XScale but kernel config is for arm926
family. 

So does it it XScale or arm926? Instead of doing optimalizations by
hand it should require proper conf/machine/include/tune-* config (which 
also set proper PACKAGE_ARCH instead of plain 'arm' one).

But there is a problem with arm926 family. Or rather GCC has problem with
it. In gcc 3.x it was '-mtune=arm926ejs' but in gcc 4.x it is 
'-mtune=arm926ej-s'.

As effect I prefer to leave fixing config to someone who is familiar with
this machine as I do not know does they use gcc 3.x or gcc 4.x for their
development.

I will skip this machine in my future ARM testing until this config will
get fixed.

It use u-boot but does not provide machine config for it. 

Kernel build fails:

arch/arm/mach-at91sam9260/devices.c:43: error: static declaration of 'dbgu_resources' follows non-static declaration
include/asm/arch/board.h:37: error: previous declaration of 'dbgu_resources' was here
make[1]: *** [arch/arm/mach-at91sam9260/devices.o] Error 1

Full log: http://ewi546.ewi.utwente.nl/tinderbox/showlog.pl?machine_id=177&logfile=20070108132854.log

=============================================================================

omap1610h2 use 2.6.9 kernel which cannot be built using gcc 4.1.1:

cc1: error: unrecognized command line option "-mapcs-32"
cc1: error: unrecognized command line option "-mshort-load-bytes"

omap1510inn, omap1610h2, omap1710h3 and omap2420h4 machines in OE are not
maintained since Chris 'kergoth' Larson stopped working for TI. 

I would like to schedule them for removal.

=============================================================================

gumstix - this machine needs config cleanup:

1. not yet converted to 'task-base' format
2. has PACKAGE_ARCH set to 'arm' instead of 'armv5te' (XScale default)
3. use own optimalizations - such stuff belongs to distro config rather then
   machine one

Anyway this one builds 'task-base' but resulting rootfs will lack needed
utils.

=============================================================================

Maintainers of 'broken' machines - would be nice if you will try to fix
those problems. I will try to do such builds in future.

-- 
JID: hrw-jabber.org
OpenEmbedded developer/consultant

  Never argue with an idiot, they'll just bring you down to their level 
  and beat you with experience.

Re: Status of ARM machines in OE

[Marcin Juszkiewicz]

Dnia poniedziałek, 8 stycznia 2007 16:18, Marcin Juszkiewicz napisał:

Ops. xxs1500 is mipsel not arm - sorry.

Anyway it is broken too ;D

-- 
JID: hrw-jabber.org
OpenEmbedded developer/consultant

               Beware of Programmers who carry screwdrivers.
               		-- Leonard Brandwein

Re: Status of ARM machines in OE

[Justin Patrin]

On 1/8/07, Marcin Juszkiewicz <openembedded@hrw.one.pl> wrote:
>
> During weekend I build 'task-base' for all ARM (armv4t and better)
> machines and Ångström distribution. The list included:
>
> a1200 a780 amsdelta compulab-pxa270 devkitidp-pxa255 ep93xx h1910 h1940
> h2200 h3900 h4000 h5000 h6300 htcblueangel htcuniversal hx2000 hx4700
> ks8695 logicpd-pxa270 magician mainstone mtx-3 mx21ads mx31ads
> navman-icn330 netbook-pro nokia770 omap1510inn omap1610h2 omap1710h3
> omap2420h4 omap5912osk palmld palmtx rokr-e2 rx3000 smdk2440 xxs1500
>
> This week ARM 'broken' machines are:
>
> a1200 - lack of kernel config - bug #1766
> omap1710h3 - lack of kernel config - bug #1736
> omap2420h4 - lack of kernel config - bug #1744
> xxs1500 - unfetchable kernel - bug #1737
>

What about the Zaurii?

-- 
Justin Patrin

Re: Status of ARM machines in OE

[Koen Kooi]

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

Justin Patrin schreef:
> On 1/8/07, Marcin Juszkiewicz <openembedded@hrw.one.pl> wrote:
>> During weekend I build 'task-base' for all ARM (armv4t and better)
>> machines and Ångström distribution. The list included:
>>
>> a1200 a780 amsdelta compulab-pxa270 devkitidp-pxa255 ep93xx h1910 h1940
>> h2200 h3900 h4000 h5000 h6300 htcblueangel htcuniversal hx2000 hx4700
>> ks8695 logicpd-pxa270 magician mainstone mtx-3 mx21ads mx31ads
>> navman-icn330 netbook-pro nokia770 omap1510inn omap1610h2 omap1710h3
>> omap2420h4 omap5912osk palmld palmtx rokr-e2 rx3000 smdk2440 xxs1500
>>
>> This week ARM 'broken' machines are:
>>
>> a1200 - lack of kernel config - bug #1766
>> omap1710h3 - lack of kernel config - bug #1736
>> omap2420h4 - lack of kernel config - bug #1744
>> xxs1500 - unfetchable kernel - bug #1737
>>
> 
> What about the Zaurii?

I suppose you mean 'zauri', 'zaurus' or 'zauruses', right? Or do you insist on using
made-up words and languages instead of English or Latin?

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

iD8DBQFFomyDMkyGM64RGpERAt7AAJ4iZ5MLDgfg40dMu4ERxJ864GvhRwCffMUl
KH+VIopw34aVaqG1LPmMGPg=
=St3l
-----END PGP SIGNATURE-----

Re: Status of ARM machines in OE

[Erik Hovland]

On Mon, Jan 08, 2007 at 05:08:35PM +0100, Koen Kooi wrote:
> > What about the Zaurii?
> 
> I suppose you mean 'zauri', 'zaurus' or 'zauruses', right? Or do you insist on using
> made-up words and languages instead of English or Latin?

Seeing how malleable English is, that zaurus is a product name invented
by people who speak Japanese natively and that zaurus is not an english
word - you might want to go a little softer on the zaurus v. zauruses v.
zauri v. zaurii thing.

E

-- 
Erik Hovland
mail: erik AT hovland DOT org
web: http://hovland.org/
PGP/GPG public key available on request

Re: Status of ARM machines in OE

[Marcin Juszkiewicz]

Dnia poniedziałek, 8 stycznia 2007 17:03, Justin Patrin napisał:

> What about the Zaurii?

I did builds for Zaurus machines - all builds fine.

-- 
JID: hrw-jabber.org
OpenEmbedded developer/consultant

             I'm not being rude, you are being insignificant.

Re: Status of ARM machines in OE

[Cliff Brake]

On 1/8/07, Marcin Juszkiewicz <openembedded@hrw.one.pl> wrote:

> compulab-pxa270 - use binary driver for flash so only hardfloat
> distributions
> can use this machine:
>
> arm-angstrom-linux-gnueabi-ld: ERROR: drivers/block/cl_flash/CL_FlashDrv uses FPA instructions, whereas drivers/block/cl_flash/fdrv.o does not
> arm-angstrom-linux-gnueabi-ld: failed to merge target specific data of file drivers/block/cl_flash/CL_FlashDrv
>
> Full log:
> http://ewi546.ewi.utwente.nl/tinderbox/showlog.pl?machine_id=177&logfile=20070107145519.log

Hi Marcin,

Thanks for doing the builds -- very interesting.

The issue with this drive is it uses the old abi, so non-eabi builds
should work.  I recently added a note in the kernel bb.  Softfloat is
working fine.  Is there any easy way in angstrom to specify oabi?

Thanks,
Cliff

-- 
=======================
Cliff Brake
http://bec-systems.com

Re: Status of ARM machines in OE

[Koen Kooi]

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

Cliff Brake schreef:
> On 1/8/07, Marcin Juszkiewicz <openembedded@hrw.one.pl> wrote:
> 
>> compulab-pxa270 - use binary driver for flash so only hardfloat
>> distributions
>> can use this machine:
>>
>> arm-angstrom-linux-gnueabi-ld: ERROR: drivers/block/cl_flash/CL_FlashDrv uses FPA instructions, whereas drivers/block/cl_flash/fdrv.o does not
>> arm-angstrom-linux-gnueabi-ld: failed to merge target specific data of file drivers/block/cl_flash/CL_FlashDrv
>>
>> Full log:
>> http://ewi546.ewi.utwente.nl/tinderbox/showlog.pl?machine_id=177&logfile=20070107145519.log
> 
> Hi Marcin,
> 
> Thanks for doing the builds -- very interesting.
> 
> The issue with this drive is it uses the old abi, so non-eabi builds
> should work.  

only (soft)fpa, not (soft)vfp.

> I recently added a note in the kernel bb.  Softfloat is
> working fine.  Is there any easy way in angstrom to specify oabi?

DISTRO="angstrom-2007.1-oabi"

regards,

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

iD8DBQFFom/hMkyGM64RGpERArsZAJ9QGPn7EUooCbWNXl0bmFzBFQqjzgCeLIjU
97Zxmzl7+Gn2KYOu6Va+N3I=
=hs8w
-----END PGP SIGNATURE-----

Re: Status of ARM machines in OE

[Richard Purdie]

On Mon, 2007-01-08 at 17:22 +0100, Koen Kooi wrote:
> only (soft)fpa, not (soft)vfp.
> 
> > I recently added a note in the kernel bb.  Softfloat is
> > working fine.  Is there any easy way in angstrom to specify oabi?
> 
> DISTRO="angstrom-2007.1-oabi"

Perhaps its time to introduce some kind of marking in the machine.conf
files like:

COMPATIBLE_ARM_ABI = "(oabi|eabi)" which angstrom can then error on
accordingly?

Richard

Re: Status of ARM machines in OE

[Koen Kooi]

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

Richard Purdie schreef:
> On Mon, 2007-01-08 at 17:22 +0100, Koen Kooi wrote:
>> only (soft)fpa, not (soft)vfp.
>>
>>> I recently added a note in the kernel bb.  Softfloat is
>>> working fine.  Is there any easy way in angstrom to specify oabi?
>> DISTRO="angstrom-2007.1-oabi"
> 
> Perhaps its time to introduce some kind of marking in the machine.conf
> files like:
> 
> COMPATIBLE_ARM_ABI = "(oabi|eabi)" which angstrom can then error on
> accordingly?

s/error on/adjust abi accordingly/ ideally.

Note that you can run eabi on strongarm by forcing -no-thumb-interwork, which requires
some glibc hacking to remove -Werror from various makefiles. This doesn't buy you much,
since you still need seperate feeds for armv4, armv4t, armv5te and armv6 (and powerpc,
x86, etc).

regards,

Koen

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

iD8DBQFFookuMkyGM64RGpERAnAiAJ9XK6nh97vfWaeOkoAqQannDB/h3wCgnvNs
5Q3+GJMeI/oD++cJEHpfxZk=
=U+WC
-----END PGP SIGNATURE-----

Re: Status of ARM machines in OE

[Paul Sokolovsky]

Hello Koen,

Monday, January 8, 2007, 8:10:54 PM, you wrote:

[]

> Note that you can run eabi on strongarm by forcing
> -no-thumb-interwork, which requires
> some glibc hacking to remove -Werror from various makefiles. This doesn't buy you much,
> since you still need seperate feeds for armv4, armv4t, armv5te and armv6 (and powerpc,
> x86, etc).

  But that's not an issue, as Angstrom  already maintains bunch of
feeds, and adding "armv4" won't be lots of trouble, right?

  So, I just hope that SA machine people (or someone else) will hack
in EABI support eventually to gcc/glibc/etc. ...

> regards,

> Koen


-- 
Best regards,
 Paul                            mailto:pmiscml@gmail.com

Re: Status of ARM machines in OE

[Koen Kooi]

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

Paul Sokolovsky schreef:
> Hello Koen,
> 
> Monday, January 8, 2007, 8:10:54 PM, you wrote:
> 
> []
> 
>> Note that you can run eabi on strongarm by forcing
>> -no-thumb-interwork, which requires
>> some glibc hacking to remove -Werror from various makefiles. This doesn't buy you much,
>> since you still need seperate feeds for armv4, armv4t, armv5te and armv6 (and powerpc,
>> x86, etc).
> 
>   But that's not an issue, as Angstrom  already maintains bunch of
> feeds, and adding "armv4" won't be lots of trouble, right?

We're going to do it a little different:

PACKAGE_ARCH = "arm-oabi"
PACKAGE_EXTRA_ARCHS = "arm-oabi"

That makes the angstrom EABI and OABI feeds incompatible for the package manager.

regards,

Koen

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

iD8DBQFFoo7RMkyGM64RGpERAn3HAJ9mkiJbmd4yY2NE3oWbj5kunwINnwCff1Fj
FCmFOnxF86/8hpM3IqzV88k=
=jV5u
-----END PGP SIGNATURE-----

Re: Status of ARM machines in OE

[Marcin Juszkiewicz]

Dnia poniedziałek, 8 stycznia 2007 17:19, Cliff Brake napisał:
> > compulab-pxa270 - use binary driver for flash so only hardfloat
> > distributions can use this machine:

> Thanks for doing the builds -- very interesting.
>
> The issue with this drive is it uses the old abi, so non-eabi builds
> should work.  I recently added a note in the kernel bb.  Softfloat is
> working fine.  Is there any easy way in angstrom to specify oabi?

Yes - DISTRO="angstrom-2007.1-oabi" makes it. I will do such build soon 
for armv4 and compulab-pxa270 machines.

-- 
JID: hrw-jabber.org
OpenEmbedded developer/consultant

              A person who aims at nothing is sure to hit it.

Re: Status of ARM machines in OE

[Marcin Juszkiewicz]

Dnia poniedziałek, 8 stycznia 2007 20:34, Marcin Juszkiewicz napisał:

> > Is there any easy way in angstrom to specify oabi?
>
> Yes - DISTRO="angstrom-2007.1-oabi" makes it. I will do such build soon
> for armv4 and compulab-pxa270 machines.

I did "angstrom-2007.1-oabi" build for collie, h3600, h3800, simpad,
compulab-pxa270. Results:

collie, h3600, compulab-pxa270: built

h3800, simpad: broke in same way as h1910/htcbluengel/rx3000 in previous run

Someone who maintain linux-handhelds kernel need this patch applied:

[hrw@misc1 org.openembedded.dev]$ mtn dif packages/linux/linux-handhelds-2.6.inc
#
# old_revision [6a7ee07ba76c72b981330c44ab8b4d0ab285b2da]
#
# patch "packages/linux/linux-handhelds-2.6.inc"
#  from [7b2de315c55f5138116f295bfab2563cbd48e6e9]
#    to [1feef0116d4e092c87c44d63f6925be52126cf9c]
#
============================================================
--- packages/linux/linux-handhelds-2.6.inc      7b2de315c55f5138116f295bfab2563cbd48e6e9
+++ packages/linux/linux-handhelds-2.6.inc      1feef0116d4e092c87c44d63f6925be52126cf9c
@@ -1,9 +1,9 @@ COMPATIBLE_HOST = "arm.*-linux"
 SECTION = "kernel"
 DESCRIPTION = "handhelds.org Linux kernel 2.6 for PocketPCs and other consumer handheld devices."
 LICENSE = "GPL"

 COMPATIBLE_HOST = "arm.*-linux"
-COMPATIBLE_MACHINE ?= '(asus730|h1910|h2200|h3600|h3900|h4000|h5000|htcblueangel|htcuniversal|hx4700|jornada56x|magician|simpad)'
+COMPATIBLE_MACHINE ?= '(asus730|h1910|h2200|h3600|h3900|h4000|h5000|htcblueangel|htcuniversal|hx4700|jornada56x|magician|simpad|rx3000|h3800)'

 # SRC_URI *must* be overriden in includer, but this is a good reference
 SRC_URI ?= "${HANDHELDS_CVS};module=linux/kernel26;tag=${@'K' + bb.data.getVar('PV',d,1).replace('.', '-')} \

-- 
JID: hrw-jabber.org
OpenEmbedded developer/consultant

          As a computer, I find your faith in technology amusing.

Re: Status of ARM machines in OE

[Koen Kooi]

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

Marcin Juszkiewicz schreef:
> Dnia poniedziałek, 8 stycznia 2007 20:34, Marcin Juszkiewicz napisał:
> 
>>> Is there any easy way in angstrom to specify oabi?
>> Yes - DISTRO="angstrom-2007.1-oabi" makes it. I will do such build soon
>> for armv4 and compulab-pxa270 machines.
> 
> I did "angstrom-2007.1-oabi" build for collie, h3600, h3800, simpad,
> compulab-pxa270. Results:
> 
> collie, h3600, compulab-pxa270: built
> 
> h3800, simpad: broke in same way as h1910/htcbluengel/rx3000 in previous run
> 
> Someone who maintain linux-handhelds kernel need this patch applied:

I don't maintain it, but added the line you pasted on irc anyway.

regards,

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

iD8DBQFFo5ngMkyGM64RGpERAj7TAJ9RMxZNVf5oSpqAF5XF/6Hjv+PpjACcCUf1
wYhM8rcYhj24CLOsQ8QtneQ=
=zuGD
-----END PGP SIGNATURE-----

Re: Status of ARM machines in OE

[Rolf Leggewie]

Marcin Juszkiewicz wrote:

> I did "angstrom-2007.1-oabi" build for collie, h3600, h3800, simpad,
> compulab-pxa270. Results:

Do I see signs of collie being supported in angstrom?

Re: Status of ARM machines in OE

[Marcin Juszkiewicz]

Dnia wtorek, 9 stycznia 2007 23:09, Rolf Leggewie napisał:
> > I did "angstrom-2007.1-oabi" build for collie, h3600, h3800, simpad,
> > compulab-pxa270. Results:
>
> Do I see signs of collie being supported in angstrom?

If collie will meet common rules then yes. 

Common rules:

- 2.6 kernel
- atleast 2 machine mentors in Angstrom
- usable for users

Currently collie meet only first...

-- 
JID: hrw-jabber.org
OpenEmbedded developer/consultant

    Progress (n.): The process through which the Internet has evolved 
    from smart people in front of dumb terminals to dumb people 
    in front of smart terminals.                          [obscurity]

Re: Status of ARM machines in OE

[Rolf Leggewie]

Marcin Juszkiewicz wrote:

>> Do I see signs of collie being supported in angstrom?
> 
> If collie will meet common rules then yes. 
> 
> Common rules:
> 
> - 2.6 kernel

but still no SD, right?

> - atleast 2 machine mentors in Angstrom
> - usable for users
> 
> Currently collie meet only first...

What is the status on the last two?  What is necessary to become machine 
mentor?  What is not usable for users in collie angstrom?

Re: Status of ARM machines in OE

[Marcin Juszkiewicz]

Dnia środa, 10 stycznia 2007 14:00, Rolf Leggewie napisał:
> Marcin Juszkiewicz wrote:
> >> Do I see signs of collie being supported in angstrom?
> >
> > If collie will meet common rules then yes.
> >
> > Common rules:
> >
> > - 2.6 kernel
>
> but still no SD, right?

What does rule 'Machine must have working 2.6 kernel' has to do with SD? 
There are machines which run Angstrom which does not have SD slot.

> > - atleast 2 machine mentors in Angstrom
> > - usable for users
> >
> > Currently collie meet only first...

> What is the status on the last two?  

> What is necessary to become machine mentor? 

Contact Angstrom maintainers, test machine images, help fixing bugs and 
improve machine support.

> What is not usable for users in collie angstrom? 

There is no MMC/SD support, few other things are not finished in 2.6 port 
too. Before those get fixed collie is unusable for users.

-- 
JID: hrw-jabber.org
OpenEmbedded developer/consultant

           Despair is just another word for 'not enough coffee' 
           		Wally

Re: Status of ARM machines in OE

[Rolf Leggewie]

Marcin Juszkiewicz wrote:

>> but still no SD, right?
> 
> What does rule 'Machine must have working 2.6 kernel' has to do with SD? 

That rule (which I understand and support fully, BTW) will break SD on 
the collie.

>> What is not usable for users in collie angstrom? 
> 
> There is no MMC/SD support, few other things are not finished in 2.6 port 
> too. Before those get fixed collie is unusable for users.

Well SD is certainly a major drawback.  If that is the most important 
thing, I would not call that "unusable", though.  Just certain parts are 
unusable.

Any machine maintainers for collie, yet, or nobody at all?

Re: Status of ARM machines in OE

[Paul Sokolovsky]

Hello Marcin,

Wednesday, January 10, 2007, 12:22:34 AM, you wrote:

> Dnia wtorek, 9 stycznia 2007 23:09, Rolf Leggewie napisał:
>> > I did "angstrom-2007.1-oabi" build for collie, h3600, h3800, simpad,
>> > compulab-pxa270. Results:
>>
>> Do I see signs of collie being supported in angstrom?

> If collie will meet common rules then yes. 

> Common rules:

> - 2.6 kernel
> - atleast 2 machine mentors in Angstrom

  Now, 2? Isn't that too steep? It's understood when 2 or more people
required to support some big software system, but what about machines?
Such requirement is simply not realistic for most of them. Of HH.org
machines, only hx4700 would qualify fully. Even for h2200 that would
be questionable, but thanks god, Koen has it, so assuming Matt Reimer
mentors it, it is safe.

  With such rules, there won't be Linux for PDA any time soon, only
Linux for Zauruses, Linux for iPaq hx4700, and other toy Linuxes.

  What's really required here is a point of view change. It would be
ludicrous to think that Debian would require two maintainers for each
motherboard model of each vendor. Debian simply doesn't deal with such
things as motherboards. It supports specific CPU architectures, that's
enough. And at the same time Linux distros for PDAs are obsessed over
PDA motherboards. There's even animism tendencies considering that
"machine" is something more than a tag given to some motherboard
implementation of the architecture, but also a bootloader spirit lives
there, and by conjuring it away one can beshrew rage of gods, etc.,
etc.

  So, stepping away from that towards architecture implementations as
such, I may offer myself as a mentor for all handhelds.org-maintained
machines. In 2.6 kernels, we have all the needed commonality and
generality (and work towards more) to treat all our machines
consistently. So, what's really required is someone to physically test
boot on a machine, and that's enough. Not boot means upstream (hh.org)
kernel problem, otherwise, it works (bugs are either common or due to
specific kernel bug also).



-- 
Best regards,
 Paul                            mailto:pmiscml@gmail.com

Re: Status of ARM machines in OE

[Koen Kooi]

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

Paul Sokolovsky schreef:

<snip>

> So, what's really required is someone to physically test
> boot on a machine, and that's enough. Not boot means upstream (hh.org)
> kernel problem, otherwise, it works (bugs are either common or due to
> specific kernel bug also).

That's not true. Yesterday my h5550 didn't boot, but I got it booting by using the a
different /boot/params and fixing the fstab.

regards,

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

iD8DBQFFpPp5MkyGM64RGpERAufPAJ9zrilcygvsv37cQWjCfctF1E55AQCgjc2g
ZuLEFeSPq1rtpWFOM7don4U=
=XWQM
-----END PGP SIGNATURE-----

Re: [Angstrom-devel] Status of ARM machines in OE

[Paul Sokolovsky]

Hello Koen,

Wednesday, January 10, 2007, 4:38:49 PM, you wrote:

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

> Paul Sokolovsky schreef:

> <snip>

>> So, what's really required is someone to physically test
>> boot on a machine, and that's enough. Not boot means upstream (hh.org)
>> kernel problem, otherwise, it works (bugs are either common or due to
>> specific kernel bug also).

> That's not true. Yesterday my h5550 didn't boot, but I got it booting by using the a
> different /boot/params and fixing the fstab.

  Yes, I know, that what I wrote to the large part is wishful thinking
at this time, but if people indeed change point of view to such
discrepancies from "we're stuck at that" to "temporary evil", it would
quickly become reality.

  As for h5000 hardships, recent bootloader talk was exactly about
that - if all or none of pocketpc's used /boot/params, there woild be much
less probability for such issues, and they would reoccur much rarely
and be fixed much quicker.

  fstab is of course more hard case - while it's very tempting to
reduce to a single standard MTD layout, it's less practical that
having single generic bootloader support (+ any number of adhoc
model-specific ones, of course).

> regards,

> Koen


-- 
Best regards,
 Paul                            mailto:pmiscml@gmail.com

Re: Status of ARM machines in OE

[pHilipp Zabel]

On 1/8/07, Marcin Juszkiewicz <openembedded@hrw.one.pl> wrote:
>
> magician is OK - builds without problem.

That is nice to hear. I'll throw away my months old gpe image and
start to build a fresh one to see wether it only builds, or even
actually works  without problems :)

> But as this is HTC Magician then
> maybe we should rename it to htcmagician to follow other htc phones
> (htcblueangel, htcuniversal) which are in OE?

Please no, I wouldn't want hx4700 to be renamed to hpipaqhx4700,
and pretty much for the same reasons I strongly prefer the name
magician. It is easier to type and easier on the eyes.

regards
Philipp

Re: Status of ARM machines in OE

[Paul Sokolovsky]

Hello Marcin,

Monday, January 8, 2007, 5:18:46 PM, you wrote:

> During weekend I build 'task-base' for all ARM (armv4t and better) 
> machines and Angstrom distribution.

  Thanks!

> The list included:

[]

> h1910 - kernel broken:
> arch/arm/mach-pxa/built-in.o: In function `h4000_set_led':
> dma_needs_bounce.c:(.text+0x1ca0): undefined reference to `ipaq_asic3_set_led'
> dma_needs_bounce.c:(.text+0x1cd0): undefined reference to `ipaq_asic3_set_led'
> dma_needs_bounce.c:(.text+0x1ce4): undefined reference to `ipaq_asic3_set_led'
> dma_needs_bounce.c:(.text+0x1d04): undefined reference to `ipaq_asic3_set_led'
> dma_needs_bounce.c:(.text+0x1d18): undefined reference to `ipaq_asic3_set_led'
> make: *** [.tmp_vmlinux1] Error 1

> Full log: 
> http://ewi546.ewi.utwente.nl/tinderbox/showlog.pl?machine_id=177&logfile=20070107150654.log

> =============================================================================

> htcblueangel - kernel broken is same place as h1910

> =============================================================================

> rx3000 - kernel broken is same place as h1910

  Well, all these 3 actually "no defconfig" cases, with some "default"
one being used, I'll take care of this. But htcblueangel and rx3000
are more complex, they are not supported by Angstrom formally, and
Angstrom's default linux-handhelds-2.6 version doesn't work for them,
they require much more recent versions. I'm going to add ?= version
spec to machine config, this way it will be fully overridable by
local.conf/distro.


> =============================================================================

> htcuniversal - kernel broken:

CONFIG_AUDIT is broken in latest versions, old defconfig... Will be
taken care of.

[]

> =============================================================================

> magician is OK - builds without problem. But as this is HTC Magician then
> maybe we should rename it to htcmagician to follow other htc phones
> (htcblueangel, htcuniversal) which are in OE?

  Philipp, please, please!!! ;-)))

  Generally, we should start to think about machine naming convention.
As an informal proposal, I would like to ask machine maintainers
considering adopting naming scheme where device name/model is
prepended with vendor or well-known brand ID. "htc*" devices are good
example of this.

  iPaqs also used to be, starting with well-known "h" prefix ;-). But
no longer, with series have expanded. So, they'd ideally be called
like "ipaq-hx4700", etc. Yes, that applies to Zaurus too ;-).

  So again, I don't think it's good time to set some rules right now,
but sooner or later OE will need them. So, it would be nice to start
exchanging ideas now, and one of them is above ;-).


> =============================================================================

> Maintainers of 'broken' machines - would be nice if you will try to fix
> those problems. I will try to do such builds in future.

  Such regression builds are very good thing. It would be nice if
there was some background info available about them - are they
automated and to what extent, how often they are run, etc. I
understand that at this time that info will be likely "this is test,
pilot run", but if you have some plans/ideas regarding that, please
share them.



-- 
Best regards,
 Paul                            mailto:pmiscml@gmail.com