[RFC] Make some big but only formal change before next stable - SRCPV

[RFC] Make some big but only formal change before next stable - SRCPV

[Martin Jansa]

Hi,

what it missing in bitbake master to call it 1.9.0 and switch all to
this version with sanity check? I'm using master on my builder and SHR
builder and it works good. (Only issue I still have is oestats-client -
reported yesterday to bitbake-dev list).

In this version is support for LOCALCOUNT override, which makes SRCPV
usable with multiple builders with shared feeds. Which was only big
complaint against using SRCPV for git recipes.
http://git.openembedded.org/cgit.cgi/bitbake/commit/?id=2f32735463159e9e42e03819d6b505dba49c7f17

My SRCPV migration branch is quickly rotting.. but I'm willing to
check/do it again, because resolving conflicts with every small change
in git recipe takes me more time than merging SRCPV to oe.dev would
take.

SRCPV migration thread:
http://www.mail-archive.com/openembedded-devel@lists.openembedded.org/msg01325.html
or really long and boring description 
http://www.mail-archive.com/openembedded-devel@lists.openembedded.org/msg01607.html

Thanks and cheers,

-- 
uin:136542059                jid:Martin.Jansa@gmail.com
Jansa Martin                 sip:jamasip@voip.wengo.fr 
JaMa                         

Re: [RFC] Make some big but only formal change before next stable - SRCPV

[Phil Blundell]

On Tue, 2010-03-09 at 08:53 +0100, Martin Jansa wrote:
> My SRCPV migration branch is quickly rotting.. but I'm willing to
> check/do it again, because resolving conflicts with every small change
> in git recipe takes me more time than merging SRCPV to oe.dev would
> take.

At the risk of sounding like a terrible dimwit, I am still not entirely
clear on why you can't just do something like

PKGV = "0.0+git${@bb.fetch.get_srcrev(d)}.${PV}"

which seems like it would make this whole problem go away without
needing any further changes to bitbake.

p.

Re: [RFC] Make some big but only formal change before next stable - SRCPV

[Martin Jansa]

On Tue, Mar 09, 2010 at 10:36:28AM +0000, Phil Blundell wrote:
> On Tue, 2010-03-09 at 08:53 +0100, Martin Jansa wrote:
> > My SRCPV migration branch is quickly rotting.. but I'm willing to
> > check/do it again, because resolving conflicts with every small change
> > in git recipe takes me more time than merging SRCPV to oe.dev would
> > take.
> 
> At the risk of sounding like a terrible dimwit, I am still not entirely
> clear on why you can't just do something like

Feeling the same here now..

> PKGV = "0.0+git${@bb.fetch.get_srcrev(d)}.${PV}"

How this should work?

is PV at the end typo from PR or do you really mean that srcrev has higher
priority than PV?

> which seems like it would make this whole problem go away without
> needing any further changes to bitbake.

All changes are already there. What's new in master is optional support
for forcing that incrementing part to same value across builders, without
local cache synchronization.

And also caching for revs returned when using BB_GIT_CLONE_FOR_SRCREV,
which can improve performance instead of running "git rev-list | wc -l"
with every get_srcrev(d).

Regards,

-- 
uin:136542059                jid:Martin.Jansa@gmail.com
Jansa Martin                 sip:jamasip@voip.wengo.fr 
JaMa                         

Re: [RFC] Make some big but only formal change before next stable - SRCPV

[Phil Blundell]

On Tue, 2010-03-09 at 12:59 +0100, Martin Jansa wrote:
> On Tue, Mar 09, 2010 at 10:36:28AM +0000, Phil Blundell wrote:
> > On Tue, 2010-03-09 at 08:53 +0100, Martin Jansa wrote:
> > > My SRCPV migration branch is quickly rotting.. but I'm willing to
> > > check/do it again, because resolving conflicts with every small change
> > > in git recipe takes me more time than merging SRCPV to oe.dev would
> > > take.
> > 
> > At the risk of sounding like a terrible dimwit, I am still not entirely
> > clear on why you can't just do something like
> 
> Feeling the same here now..
> 
> > PKGV = "0.0+git${@bb.fetch.get_srcrev(d)}.${PV}"
> 
> How this should work?
> 
> is PV at the end typo from PR or do you really mean that srcrev has higher
> priority than PV?

Yes, I would have thought you would want the srcrev to have higher
priority than PV.  In fact, you probably don't even want to have PV in
there at all.  You certainly wouldn't want PR in there since that goes
in ${PKGR}.

p.

Re: [RFC] Make some big but only formal change before next stable - SRCPV

[Koen Kooi]

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

On 09-03-10 08:53, Martin Jansa wrote:
> Hi,
> 
> what it missing in bitbake master to call it 1.9.0 and switch all to
> this version with sanity check? I'm using master on my builder and SHR
> builder and it works good. (Only issue I still have is oestats-client -
> reported yesterday to bitbake-dev list).
> 
> In this version is support for LOCALCOUNT override, which makes SRCPV
> usable with multiple builders with shared feeds. Which was only big
> complaint against using SRCPV for git recipes.
> http://git.openembedded.org/cgit.cgi/bitbake/commit/?id=2f32735463159e9e42e03819d6b505dba49c7f17

What I had plan to RFC, but forgot to, is to move the count vars out of
TMPDIR. My PV suddenly went backwards for a git recipe I'm using,
because I removed TMPDIR.

regards,

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

iD8DBQFLllR4MkyGM64RGpERAlgWAJ4l4Ntky6O7MT62bvsOCHWWy+0+iQCfUrpI
X9Yw40O4OnhSuZfi6K40oTY=
=BSSV
-----END PGP SIGNATURE-----

Re: [RFC] Make some big but only formal change before next stable - SRCPV

[Phil Blundell]

On Tue, 2010-03-09 at 15:00 +0100, Koen Kooi wrote:
> What I had plan to RFC, but forgot to, is to move the count vars out of
> TMPDIR. My PV suddenly went backwards for a git recipe I'm using,
> because I removed TMPDIR.

That does rather sound like a further argument for just not using the
local count variables at all and just figuring out a monotonic count
from the SCM (using git-rev-list or whatever) at packaging time.

p.

Re: [RFC] Make some big but only formal change before next stable - SRCPV

[Sebastian Spaeth]

On Tue, 09 Mar 2010 20:41:59 +0000, Phil Blundell <philb@gnu.org> wrote:
> That does rather sound like a further argument for just not using the
> local count variables at all and just figuring out a monotonic count
> from the SCM (using git-rev-list or whatever) at packaging time.

That sounds rather nice, however I am not sure if git easily gives you
such a number. I know that mercurial does.

spaetz

Re: [RFC] Make some big but only formal change before next stable - SRCPV

[Koen Kooi]

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

On 10-03-10 10:48, Sebastian Spaeth wrote:
> On Tue, 09 Mar 2010 20:41:59 +0000, Phil Blundell <philb@gnu.org> wrote:
>> That does rather sound like a further argument for just not using the
>> local count variables at all and just figuring out a monotonic count
>> from the SCM (using git-rev-list or whatever) at packaging time.
> 
> That sounds rather nice, however I am not sure if git easily gives you
> such a number. I know that mercurial does.

Both are local and don't handle rebasing well.

regards,

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

iD8DBQFLl7MqMkyGM64RGpERAriNAJ9fixaWg0G7NHkEKv/UOuGXbJYwIQCeNmxU
Oh51RpHDi6I3D4q+IzbPbYo=
=lB78
-----END PGP SIGNATURE-----