OE.dev roadworks notification

OE.dev roadworks notification

[Richard Purdie]

Hi,

At OEDEM last year and in various other discussions it was decided that
sysroot support was in OE's interests. Since then I successfully rolled
that out in Poky, I'm now in a position to do the same for OE.dev.

On a similar note, I'll be doing some work on packaged staging over the
next few weeks. Again, its been agreed that its in the interests of
OE.dev on many occasions and now I/we have a chance for action which I'm
planing to make good use of it. 

Both sets of changes have the potential to break things. I will
obviously try to avoid that and endeavour to fix any breakage that does
happen as quickly as possible. I'd just ask that people keep an eye on
the commit messages and mailing list and appreciate that there may be
some instability. If your project/use of OE can't cope with the
instability and time debugging anything, just don't update or use the
stable branch!

With packaged staging one thing we need to decide whether ultimately its
something optional which people opt into, or whether having staging
under control of a package manager is something desirable at all times.
My gut feeling is that its something we should really make mandatory or
opt-out, maybe not at first but certainly longer term but I'm open to
views on that.

Note that I'm not going to RFC everything since the volume of changes
will mean we'd never get anywhere. I think the core team and other devs
trust me enough not to do anything too insane. If anyone does see
anything they don't like by all means question it on the mailing list.

As an opening salvo, I'll be updating the minimum bitbake version
requirement to 1.8.10 and removing all the compatibility code we have to
do with copyfile and file locking (packaged staging will need the
locking functions added in 1.8.10).

Cheers,

Richard

Re: OE.dev roadworks notification

[Graeme Gregory]

On Tue, 19 Feb 2008 11:14:52 +0000
Richard Purdie <rpurdie@rpsys.net> wrote:
> With packaged staging one thing we need to decide whether ultimately
> its something optional which people opt into, or whether having
> staging under control of a package manager is something desirable at
> all times. My gut feeling is that its something we should really make
> mandatory or opt-out, maybe not at first but certainly longer term
> but I'm open to views on that.
> 
My feeling is opt-out. That way you have to read the docs to get out
and that means you have to understand why you are swimming upstream.

One of the most common problems in OE for beginners seems to be
building packages that conflict in staging. Having the no breaking
system be the default would be the best.

Graeme (XorA)

Re: OE.dev roadworks notification

[Jeremy Lainé]

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


> My feeling is opt-out. That way you have to read the docs to get
> out and that means you have to understand why you are swimming
> upstream.
>
> One of the most common problems in OE for beginners seems to be
> building packages that conflict in staging. Having the no breaking
> system be the default would be the best.
I would also prefer to see packaged staging as the default.
It will go a long way towards making builds deterministic as
opposed to the current situation where you may or may not
have cruft leftover in your staging directory, depending on
what you built in the past.

- --
Jeremy LAINE
Bolloré telecom | 11bis, rue Scribe | F-75009 Paris
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHusZpCL1n3lVB9EYRApO1AJ4lgLwI5o9dVhYEgDMTAtiYv8UJJwCcCuF2
Bo2NZKJ5CNw/9TdCaIanuUQ=
=8yhs
-----END PGP SIGNATURE-----

Re: OE.dev roadworks notification

[Junqian Gordon Xu]

On 02/19/2008 06:14 AM, Richard Purdie wrote:
> On a similar note, I'll be doing some work on packaged staging over the
> next few weeks. Again, its been agreed that its in the interests of
> OE.dev on many occasions and now I/we have a chance for action which I'm
> planing to make good use of it. 

The following bug (typo) was detected only with INHERIT += 
"packaged-staging2", which is good, however the error message is too 
misleading to decipher. As you are on it, you may want to take a look to 
see if there is a way to provide a saner error message.

http://lists.linuxtogo.org/pipermail/openembedded-commits/2008-January/011276.html

Regards
Gordon

Re: OE.dev roadworks notification

[Junqian Gordon Xu]

On 02/19/2008 06:14 AM, Richard Purdie wrote:
> On a similar note, I'll be doing some work on packaged staging over the
> next few weeks. Again, its been agreed that its in the interests of
> OE.dev on many occasions and now I/we have a chance for action which I'm
> planing to make good use of it. 

The following bug (typo) was detected only with INHERIT += 
"packaged-staging2", which is good, however the error message is too 
misleading to decipher. As you are on it, you may want to take a look to 
see if there is a way to provide a saner error message.

http://lists.linuxtogo.org/pipermail/openembedded-commits/2008-January/011276.html

Regards
Gordon

Re: OE.dev roadworks notification

[Leon Woestenberg]

Hello,

On Tue, Feb 19, 2008 at 1:07 PM, Jeremy Lainé
<jeremy.laine@bolloretelecom.eu> wrote:
>  I would also prefer to see packaged staging as the default.
>  It will go a long way towards making builds deterministic as
>  opposed to the current situation where you may or may not
>  have cruft leftover in your staging directory, depending on
>  what you built in the past.
>
I agree with Jeremy's reasoning.

The only deterministic build currently is a build-from-scratch.
Packages configure themselves against anything that you might have
staged but what does not end up in the image.

To what stage is .stable synchronized with current .dev? I would like
to move all my running projects across both branches so that I can
both
(1) continue working on them
(2) test Richard's efforts on .dev

And Richard, your work is greatly applauded!

Regards from a mere handful-of-package maintainer,
-- 
Leon

Re: OE.dev roadworks notification

[Koen Kooi]

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

Leon Woestenberg schreef:

| To what stage is .stable synchronized with current .dev?

People propose revisions from .dev to get backported and 2 core
developers have to ACK them to get applied, and BACKPORTS.txt[1] is
updated with those revisions. That way the stable branch is a 'subset'
of .dev, and we aim to keep 'branch-only' fixes to an absoulute minimum
('zero' is our goal).

| I would like
| to move all my running projects across both branches so that I can
| both
| (1) continue working on them
| (2) test Richard's efforts on .dev

That shouldn't be a problem, if you find something missing, apply the
change locally or propose it for inclusion.

regards,

Koen

[1] e.g.
http://www.openembedded.org:1081/revision/file/4ade39d8fb840c136daf5a450ece76d6b7280dea/BACKPORTS.txt
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)

iD8DBQFHvUsIMkyGM64RGpERAv4ZAKCZOfIdj6ykpiRh/pOo5bul3yCFOgCfSXCt
gkMYa6UTiNYgEnOoZHiIM8g=
=8o1N
-----END PGP SIGNATURE-----

Re: OE.dev roadworks notification

[Leon Woestenberg]

Hello Koen,

On Thu, Feb 21, 2008 at 10:57 AM, Koen Kooi
<koen@dominion.kabel.utwente.nl> wrote:
>
>  Leon Woestenberg schreef:
>
>  | To what stage is .stable synchronized with current .dev?
>
>  People propose revisions from .dev to get backported and 2 core
>  developers have to ACK them to get applied, and BACKPORTS.txt[1] is

Thanks. Now that you explained it, I remember you already explained
this policy earlier.

I should really RTFB (Retrain the fscking brain).

Regards,
-- 
Leon