native preferences

native preferences

[Florian Boor]

Hi all,

I tried to build a "minimal" distribution image lately and ran into some trouble
 because the old-style native packages seem to be preferred over the new ones
using BBCLASSEXTEND.
In this case we have glib-2.0-native 2.24.1 provided by the new mechanism and
2.22.1 gets build which is the latest old-style native bb.
To build an image we would have to set the preferences manually or drop all old
native packages. I do not see a reason why this behaviour should be desirable -
a native package should follow the same preferences rules like other packages
independent from the mechanism that provides them. Or do I miss something here?

Greetings

Florian

-- 
The dream of yesterday                  Florian Boor
is the hope of today                    Tel: +49 271-771091-15
and the reality of tomorrow.            Fax: +49 271-771091-19
[Robert Hutchings Goddard, 1904]        florian.boor@kernelconcepts.de
                                        http://www.kernelconcepts.de/en

Re: native preferences

[Koen Kooi]

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

On 26-07-10 11:42, Florian Boor wrote:
> Hi all,
> 
> I tried to build a "minimal" distribution image lately and ran into some trouble
>  because the old-style native packages seem to be preferred over the new ones
> using BBCLASSEXTEND.
> In this case we have glib-2.0-native 2.24.1 provided by the new mechanism and
> 2.22.1 gets build which is the latest old-style native bb.

That's minimals broken by design behaviour, it refuses to pin any
version. So either pin the version or switch to a sane distro like
angstrom or shr.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFMTXI1MkyGM64RGpERAi1eAJ4q5Zwx4OTwV4sfjDFQ4GBE4uClbACfd2wo
dYefl4wTR0njfF5r+bezfKs=
=SfWy
-----END PGP SIGNATURE-----

Re: native preferences

[Frans Meulenbroeks]

2010/7/26 Koen Kooi <k.kooi@student.utwente.nl>

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 26-07-10 11:42, Florian Boor wrote:
> > Hi all,
> >
> > I tried to build a "minimal" distribution image lately and ran into some
> trouble
> >  because the old-style native packages seem to be preferred over the new
> ones
> > using BBCLASSEXTEND.
> > In this case we have glib-2.0-native 2.24.1 provided by the new mechanism
> and
> > 2.22.1 gets build which is the latest old-style native bb.
>
> That's minimals broken by design behaviour, it refuses to pin any
> version. So either pin the version or switch to a sane distro like
> angstrom or shr.
>


Hm. I would expect that if there is a glib-2.0 recipe 2.24.1 with
BBCLASSEXTEND = "native" and  no pinning and no DEFAULT_PREFERENCE being in
effect, that 2.24.1 gets build and not 2.22.1 which has a real native
recipe. To me recipes with -native and implict ones with BBCLASSEXTEND =
"native" are equivalent prioritywise, so I'd say this is a bug in the class
code.

Frans

> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.5 (Darwin)
>
> iD8DBQFMTXI1MkyGM64RGpERAi1eAJ4q5Zwx4OTwV4sfjDFQ4GBE4uClbACfd2wo
> dYefl4wTR0njfF5r+bezfKs=
> =SfWy
> -----END PGP SIGNATURE-----
>
>
> _______________________________________________
> Openembedded-devel mailing list
> Openembedded-devel@lists.openembedded.org
> http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-devel
>

Re: native preferences

[Koen Kooi]

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

On 26-07-10 14:08, Frans Meulenbroeks wrote:
> 2010/7/26 Koen Kooi <k.kooi@student.utwente.nl>
> 
> On 26-07-10 11:42, Florian Boor wrote:
>>>> Hi all,
>>>>
>>>> I tried to build a "minimal" distribution image lately and ran into some
> trouble
>>>>  because the old-style native packages seem to be preferred over the new
> ones
>>>> using BBCLASSEXTEND.
>>>> In this case we have glib-2.0-native 2.24.1 provided by the new mechanism
> and
>>>> 2.22.1 gets build which is the latest old-style native bb.
> 
> That's minimals broken by design behaviour, it refuses to pin any
> version. So either pin the version or switch to a sane distro like
> angstrom or shr.
> 
> 
> 
>> Hm. I would expect that if there is a glib-2.0 recipe 2.24.1 with
>> BBCLASSEXTEND = "native" and  no pinning and no DEFAULT_PREFERENCE being in
>> effect, that 2.24.1 gets build and not 2.22.1 which has a real native
>> recipe. To me recipes with -native and implict ones with BBCLASSEXTEND =
>> "native" are equivalent prioritywise, so I'd say this is a bug in the class
>> code.

And that bug has been known for months and reported multiple times by
people attempting to use minimal. And for so far the bug has been fixed
in bitbake 1.10, but minimal still refuses to pin versions, so the
advice to switch distro still stands.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFMTX37MkyGM64RGpERArnKAJ9jd/pPEhb/XdRXTJstqJq0/5zD6ACfSToc
5nInC0ZKEPbp3x1zlBNPuDI=
=IBvv
-----END PGP SIGNATURE-----

Re: native preferences

[Dr. Michael Lauer]

Am 26.07.2010 um 14:22 schrieb Koen Kooi:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On 26-07-10 14:08, Frans Meulenbroeks wrote:
>> 2010/7/26 Koen Kooi <k.kooi@student.utwente.nl>
>> 
>> On 26-07-10 11:42, Florian Boor wrote:
>>>>> Hi all,
>>>>> 
>>>>> I tried to build a "minimal" distribution image lately and ran into some
>> trouble
>>>>> because the old-style native packages seem to be preferred over the new
>> ones
>>>>> using BBCLASSEXTEND.
>>>>> In this case we have glib-2.0-native 2.24.1 provided by the new mechanism
>> and
>>>>> 2.22.1 gets build which is the latest old-style native bb.
>> 
>> That's minimals broken by design behaviour, it refuses to pin any
>> version. So either pin the version or switch to a sane distro like
>> angstrom or shr.
>> 
>> 
>> 
>>> Hm. I would expect that if there is a glib-2.0 recipe 2.24.1 with
>>> BBCLASSEXTEND = "native" and  no pinning and no DEFAULT_PREFERENCE being in
>>> effect, that 2.24.1 gets build and not 2.22.1 which has a real native
>>> recipe. To me recipes with -native and implict ones with BBCLASSEXTEND =
>>> "native" are equivalent prioritywise, so I'd say this is a bug in the class
>>> code.
> 
> And that bug has been known for months and reported multiple times by
> people attempting to use minimal. And for so far the bug has been fixed
> in bitbake 1.10, but minimal still refuses to pin versions, so the
> advice to switch distro still stands.

Of course the preferred way - rather than switching distros - is to fix this obvious bug
in the OE metadata.

Cheers,

:M:

Re: native preferences

[Florian Boor]

Hi,

Dr. Michael Lauer schrieb:
> Of course the preferred way - rather than switching distros - is to fix this obvious bug
> in the OE metadata.

yes I see it like this. Switching to a distribution that pins more versions is
not really what I would call a fix. It is a workaround but should not become the
default way to deal with such issues in OE.
What we introduced here is clearly a bug in OE and not the distributions fault.

Greetings

Florian

-- 
The dream of yesterday                  Florian Boor
is the hope of today                    Tel: +49 271-771091-15
and the reality of tomorrow.            Fax: +49 271-771091-19
[Robert Hutchings Goddard, 1904]        florian.boor@kernelconcepts.de
                                        http://www.kernelconcepts.de/en

Re: native preferences

[Tom Rini]

Florian Boor wrote:
> Hi,
> 
> Dr. Michael Lauer schrieb:
>> Of course the preferred way - rather than switching distros - is to fix this obvious bug
>> in the OE metadata.
> 
> yes I see it like this. Switching to a distribution that pins more versions is
> not really what I would call a fix. It is a workaround but should not become the
> default way to deal with such issues in OE.
> What we introduced here is clearly a bug in OE and not the distributions fault.

Except that it's a bitbake issue, not a metadata issue, iirc (which is 
why it's fixed in the 1.10 branch, but not 1.8).

-- 
Tom Rini
Mentor Graphics Corporation

Re: native preferences

[Koen Kooi]

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

On 26-07-10 19:14, Tom Rini wrote:
> Florian Boor wrote:
>> Hi,
>>
>> Dr. Michael Lauer schrieb:
>>> Of course the preferred way - rather than switching distros - is to
>>> fix this obvious bug
>>> in the OE metadata.
>>
>> yes I see it like this. Switching to a distribution that pins more
>> versions is
>> not really what I would call a fix. It is a workaround but should not
>> become the
>> default way to deal with such issues in OE.
>> What we introduced here is clearly a bug in OE and not the
>> distributions fault.
> 
> Except that it's a bitbake issue, not a metadata issue, iirc (which is
> why it's fixed in the 1.10 branch, but not 1.8).

Yes, that's why I said "the bug has been fixed
in bitbake 1.10" earlier in the thread :)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFMTdYqMkyGM64RGpERAnN3AKCdCvwWQhKF8Ta4A4x/vpctt45XGgCcC66C
7swo93nDkqCt95K4OPWbTxY=
=ENxb
-----END PGP SIGNATURE-----