uclibc 0.9.29 recipe

uclibc 0.9.29 recipe

[Koen Kooi]

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

Hi,


I just started working on a uclibc 0.9.29 recipe, are people interested in it already, or
should I finish it before committing it (with DEFAULT_PREF = -1)?

regards,

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

iD8DBQFGRhOzMkyGM64RGpERAlBJAKC4Hekdx6iKB95Nglwt4elr/scvqQCggdgo
b0a0iWs5X06ZAMKRzYUfKlg=
=S7Cp
-----END PGP SIGNATURE-----

Re: uclibc 0.9.29 recipe

[Lorn Potter]

Koen Kooi wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Hi,
> 
> 
> I just started working on a uclibc 0.9.29 recipe, are people interested in it already, or
> should I finish it before committing it (with DEFAULT_PREF = -1)?
> 
> regards,

I would be, especially if it worked in an angstrom build.



-- 
Lorn 'ljp' Potter
Software Engineer, Systems Group, MES, Trolltech

Re: uclibc 0.9.29 recipe

[Koen Kooi]

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

Lorn Potter schreef:
> 
> Koen Kooi wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Hi,
>>
>>
>> I just started working on a uclibc 0.9.29 recipe, are people interested in it already, or
>> should I finish it before committing it (with DEFAULT_PREF = -1)?
>>
>> regards,
> 
> I would be, especially if it worked in an angstrom build.

I'll add it once I solved the linux-libc-headers problems. Machine and distro maintainers
need to provide their parts of the uclibc configs to make it usefull, though.

regards,

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

iD8DBQFGRh1yMkyGM64RGpERApzPAJ9dHfEynAkjYtpvJC6aKwcSLcF1PQCggceU
2wDlG7EN1RNSwOJ8YjfCItk=
=Un//
-----END PGP SIGNATURE-----

Re: uclibc 0.9.29 recipe

[Koen Kooi]

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

Koen Kooi schreef:
> Lorn Potter schreef:
>> Koen Kooi wrote:
>>> -----BEGIN PGP SIGNED MESSAGE-----
>>> Hash: SHA1
>>>
>>> Hi,
>>>
>>>
>>> I just started working on a uclibc 0.9.29 recipe, are people interested in it already, or
>>> should I finish it before committing it (with DEFAULT_PREF = -1)?
>>>
>>> regards,
>> I would be, especially if it worked in an angstrom build.
> 
> I'll add it once I solved the linux-libc-headers problems. 

Which is non-existant on 0.9.29 :)


> Machine and distro maintainers
> need to provide their parts of the uclibc configs to make it usefull, though.

Angstrom has a uClibc.distro and ixp4xxbe has a uClibc.machine.

To add support for your favourite machine: use uClibc.distro as .config, configure uclibc
for the target you want, save the top 2 blocks of .config (cpu and arch) to uClibc.distro
in uclibc-0.9.29/<machine/uClibc.machine

The recipe is by no means perfect, so any help is welcome.

regards,

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

iD8DBQFGRivdMkyGM64RGpERAoVVAKCCVsINF5WmH0XnnbpCF0oMo15sfACePayI
7vnmpum3r/QkizvA4HinAzc=
=R235
-----END PGP SIGNATURE-----

Re: uclibc 0.9.29 recipe

[Leon Woestenberg]

Hello Koen,

On 5/12/07, Koen Kooi <koen@dominion.kabel.utwente.nl> wrote:
> Koen Kooi schreef:
> To add support for your favourite machine: use uClibc.distro as .config,
> configure uclibc for the target you want, save the top 2 blocks of .config (cpu
> and arch) to uClibc.distro  in uclibc-0.9.29/<machine/uClibc.machine
>
> The recipe is by no means perfect, so any help is welcome.
>
As often, you laid the groundworks, which is often more important than
the make-it-perfect stage.

My hands are twitching to create a clean 0.9.29 .bb that more smartly
configures the library matching the OpenEmbedded configuration. But
let's hack the current setup first.

Who is using uClibc now? In other words, can I break stuff?

Regards,
-- 
Leon

Re: uclibc 0.9.29 recipe

[Koen Kooi]

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

Leon Woestenberg schreef:
> Hello Koen,
> 
> On 5/12/07, Koen Kooi <koen@dominion.kabel.utwente.nl> wrote:
>> Koen Kooi schreef:
>> To add support for your favourite machine: use uClibc.distro as .config,
>> configure uclibc for the target you want, save the top 2 blocks of .config (cpu
>> and arch) to uClibc.distro  in uclibc-0.9.29/<machine/uClibc.machine
>>
>> The recipe is by no means perfect, so any help is welcome.
>>
> As often, you laid the groundworks, which is often more important than
> the make-it-perfect stage.
> 
> My hands are twitching to create a clean 0.9.29 .bb that more smartly
> configures the library matching the OpenEmbedded configuration. But
> let's hack the current setup first.

It would be nice to set archicture using TARGET_ARCH, endiannes with the info from
siteinfo.bbclass and the the FPU with TARGET_FPU. I bet there are some more options that
could be autoconfigured.

regards,

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

iD8DBQFGR0MVMkyGM64RGpERAmpLAJ40KalkcxFc1oOFNSfOrMB7dypN7gCZAeg4
4C42BTmMvEcIiALtDpSx/eY=
=yp3m
-----END PGP SIGNATURE-----

Re: uclibc 0.9.29 recipe

[Leon Woestenberg]

perl 5.8.8 needs:

UCLIBC_SUSV3_LEGACY=y
UCLIBC_SUSV3_LEGACY_MACROS=y

Re: uclibc 0.9.29 recipe

[Koen Kooi]

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

Leon Woestenberg schreef:
> perl 5.8.8 needs:
> 
> UCLIBC_SUSV3_LEGACY=y
> UCLIBC_SUSV3_LEGACY_MACROS=y

Feel free to add those to (angstrom/)uClibc.distro

regards,

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

iD8DBQFGSFZIMkyGM64RGpERAm6LAJ9zXFxGitsAcWxB9MBPXT8qYQMrvwCcC6ro
XdSuqhAYD0TKlipoSA0pQJ0=
=d9jH
-----END PGP SIGNATURE-----

minimal-image.bb

[Stelios Koroneos]

Greetings to all !

Recently it was made apparent that bootstrap-image was generating images
that could not fit in flash restricted devices (<=4MB)
Koen has pushed angstrom-minimal-image which uses task-boot and as a result
generates images significally smaller.(800K in a jffs2 fs)

The only drawback into this (IMHO) is in the naming of the package.
Adding a distro name infront of an image package gives the impression that
this is distro specific and not a "generic" one.
I tried to push to the rep a more "generic" minimal-image.bb which was
almost identical to angstrom-minima-image but Koen had  dis-approved the
commit as duplicate.

In my opinion if angstrom-minimal-image is not distro specific should be
renamed to minimal-image or if it is distro specific a minimal-image.bb
should be created.

I discussed the matter with Koen on irc but we could not come to a
conlusion, so i am asking for the opinions of others on this, as i believe
is a issue which has to be solved to avoid any future problems.



Stelios S. Koroneos

Digital OPSiS - Embedded Intelligence
http://www.digital-opsis.com

Re: minimal-image.bb

[Rolf Leggewie]

Stelios Koroneos wrote:
> In my opinion if angstrom-minimal-image is not distro specific should be
> renamed to minimal-image

I agree with Stelios that the belief that angstrom-minimal-image is a 
distro-specific image is a reasonable assumption.  If that is not the 
case, I am in favor renaming.

Re: minimal-image.bb

[Stelios Koroneos]

> -----Original Message-----
> From: openembedded-devel-bounces@openembedded.org
> [mailto:openembedded-devel-bounces@openembedded.org]On Behalf Of
> Rolf Leggewie
> Sent: Sunday, May 13, 2007 5:35 PM
> To: openembedded-devel@openembedded.org
> Subject: Re: [oe] minimal-image.bb
>
>
> Stelios Koroneos wrote:
> > In my opinion if angstrom-minimal-image is not distro specific should be
> > renamed to minimal-image
>
> I agree with Stelios that the belief that angstrom-minimal-image is a
> distro-specific image is a reasonable assumption.  If that is not the
> case, I am in favor renaming.


There is a bigger issue here than the name of package and that the reason i
brought this matter up.
Angstrom without any dough is the biggest distro supported by OE, but its
not the only one.
No one should also dough that Koen has done a lot of work (and is still
doing) and credits should go to him for a lot of the things in OE, including
the angstrom-minimal-image.bb, but i believe the real issue goes beyond the
developer/distro thing.

Actions like that, give the wrong signals to the people who are looking at
OE for incorporation/support of their products(and believe me there are
*alot* that do that as we speak)
If a distro maintainer (or any OE developer in that fact) comes with a way
of handling things better for his distro that could be used as a generic
framework for others it should be done so and not discouraged.

I am firm believer in community work and that things that could benefit the
OE community should be handled as such and not as distro-specific.

Re: minimal-image.bb

[Rolf Leggewie]

Stelios Koroneos wrote:
> I am firm believer in community work and that things that could benefit the
> OE community should be handled as such and not as distro-specific.

I always thought that Angstrom was all about tearing walls down instead 
of building them up.

/me is baffled

Re: uclibc 0.9.29 recipe

[Jamie Lenehan]

On Sat, May 12, 2007 at 10:02:58PM +0200, Koen Kooi wrote:
[...]
> I'll add it once I solved the linux-libc-headers problems. Machine and distro maintainers
> need to provide their parts of the uclibc configs to make it usefull, though.

Do we really need different distro based settings? It's always been a
bit annoying that some things will build with uclibc for one build
but not another due to differing uclibc settings - things are a lot
easier if there's only a single configuration.

For now I'd like to see a

  mtn -e mv uclibc-0.9.29/angstrom/uClibc.distro uclibc-0.9.29/uClibc.distro

and then only add distro based configuration if it turns out to be
really needed by some distro. This'll also gives us a default
configuration for people using other distros.

-- 
 Jamie Lenehan <lenehan@twibble.org>

Re: uclibc 0.9.29 recipe

[Leon Woestenberg]

Hello Jamie,

On 5/16/07, Jamie Lenehan <lenehan@twibble.org> wrote:
> On Sat, May 12, 2007 at 10:02:58PM +0200, Koen Kooi wrote:

> For now I'd like to see a
>
>   mtn -e mv uclibc-0.9.29/angstrom/uClibc.distro uclibc-0.9.29/uClibc.distro
>
> and then only add distro based configuration if it turns out to be
> really needed by some distro. This'll also gives us a default
>
And that generic distro configuration would probably be a full
configuration (i.e. most options enabled) so that it would work for
most packages depending on C library functionality (or glibc
compatibility).

Distro's can optionally tune down the .distro by either:
- providing their own .distro
- mangling the resulting .config

Regards,
-- 
Leon

Re: uclibc 0.9.29 recipe

[Koen Kooi]

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

Jamie Lenehan schreef:
> On Sat, May 12, 2007 at 10:02:58PM +0200, Koen Kooi wrote:
> [...]
>> I'll add it once I solved the linux-libc-headers problems. Machine and distro maintainers
>> need to provide their parts of the uclibc configs to make it usefull, though.
> 
> Do we really need different distro based settings? 

I tried to adhere to the existing uclibc infrastructure when adding 0.9.29, but maybe it's
time to rethink and overhaul it.

> It's always been a
> bit annoying that some things will build with uclibc for one build
> but not another due to differing uclibc settings - things are a lot
> easier if there's only a single configuration.
> 
> For now I'd like to see a
> 
>   mtn -e mv uclibc-0.9.29/angstrom/uClibc.distro uclibc-0.9.29/uClibc.distro

Fine by me, could you do that?

> and then only add distro based configuration if it turns out to be
> really needed by some distro. This'll also gives us a default
> configuration for people using other distros.

And gives us a better way to keep our site-files in sync.

regards,

Koen

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

iD8DBQFGSrWoMkyGM64RGpERAnrXAJwJFcLlBdZpiwEsf1Uw8DBL9agFtgCaA0ny
EleWYKtPe+qa1D9PIgGiJps=
=ByLX
-----END PGP SIGNATURE-----

uclibc 0.9.29 & SYS_SIGLIST

[Jamie Lenehan]

Currently SYS_SIGLIST is disabled in uclibc but the site files all
say it's present. For the old uclibc version some of the
configurations enabled SYS_SIGLIST support and some didn't. We don't
really need it and the uclibc help says it's going to go away one day
so if no one objects I'll add the following site entries:

common-glibc

 bash_cv_under_sys_siglist=${bash_cv_under_sys_siglist=yes}
 bash_cv_sys_siglist=${bash_cv_sys_siglist=yes}
 ac_cv_have_decl_sys_siglist=${ac_cv_have_decl_sys_siglist=yes}

common-uclibc

 bash_cv_under_sys_siglist=${bash_cv_under_sys_siglist=no}
 bash_cv_sys_siglist=${bash_cv_sys_siglist=no}
 ac_cv_have_decl_sys_siglist=${ac_cv_have_decl_sys_siglist=no}

And remove the above entries from all the other site files.

This siglist things needs to be sorted out to build apr at least
(which is needed to build subversion).

-- 
 Jamie Lenehan <lenehan@twibble.org>