[STABLE] branch created: stable/2009

[STABLE] branch created: stable/2009

[Marcin Juszkiewicz]

Hi

I would like to announce creation of new branch: stable/2009 in 
OpenEmbedded repository. It was branched few days ago, build tested and 
got some bugfixes which were added to .dev tree (and back to 
stable/2009).

What was in build test? Distro "angstrom-2008.1", machines: qemuarm 
(armv5te), at91sam9263ek (armv5te), beagleboard (armv7a), qemux86 (i586) 
and vortex86sx (486sx). I tested building of next images: console, 
console-base, x11, beagleboard-demo for each of those machines. So far 
only glibc builds were tested, uclibc one is queued.

I also imported BitBake 1.8.12 (last release) into "stable/2009" branch 
as this is minimal stable version which works with this repository.

Feel free to clone and do tests. I will write more later.

Regards, 
-- 
JID:      hrw@jabber.org
Website:  http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz

Re: [STABLE] branch created: stable/2009

[Koen Kooi]

On 31-03-09 18:37, Marcin Juszkiewicz wrote:
>
> Hi
>
> I would like to announce creation of new branch: stable/2009 in
> OpenEmbedded repository. It was branched few days ago, build tested and
> got some bugfixes which were added to .dev tree (and back to
> stable/2009).
>
> What was in build test? Distro "angstrom-2008.1", machines: qemuarm
> (armv5te), at91sam9263ek (armv5te), beagleboard (armv7a), qemux86 (i586)
> and vortex86sx (486sx). I tested building of next images: console,
> console-base, x11, beagleboard-demo for each of those machines. So far
> only glibc builds were tested, uclibc one is queued.
>
> I also imported BitBake 1.8.12 (last release) into "stable/2009" branch
> as this is minimal stable version which works with this repository.
>
> Feel free to clone and do tests. I will write more later.

Great!

I already have a few patches queued up for the stablebranch:

http://dominion.thruhere.net/koen/OE/for-stable/

What's the proper procedure for getting those in?

regards,

Koen

Re: [STABLE] branch created: stable/2009

[Tom Rini]

On Tue, Mar 31, 2009 at 07:00:59PM +0200, Koen Kooi wrote:
> On 31-03-09 18:37, Marcin Juszkiewicz wrote:
>>
>> Hi
>>
>> I would like to announce creation of new branch: stable/2009 in
>> OpenEmbedded repository. It was branched few days ago, build tested and
>> got some bugfixes which were added to .dev tree (and back to
>> stable/2009).
>>
>> What was in build test? Distro "angstrom-2008.1", machines: qemuarm
>> (armv5te), at91sam9263ek (armv5te), beagleboard (armv7a), qemux86 (i586)
>> and vortex86sx (486sx). I tested building of next images: console,
>> console-base, x11, beagleboard-demo for each of those machines. So far
>> only glibc builds were tested, uclibc one is queued.
>>
>> I also imported BitBake 1.8.12 (last release) into "stable/2009" branch
>> as this is minimal stable version which works with this repository.
>>
>> Feel free to clone and do tests. I will write more later.
>
> Great!
>
> I already have a few patches queued up for the stablebranch:
>
> http://dominion.thruhere.net/koen/OE/for-stable/
>
> What's the proper procedure for getting those in?

And please answer in the form of a link to a wiki page :)

-- 
Tom Rini

Re: [STABLE] branch created: stable/2009

[Marcin Juszkiewicz]

Dnia wtorek, 31 marca 2009 o 19:17:27 Tom Rini napisał(a):
> > Great!
> >
> > I already have a few patches queued up for the stablebranch:
> >
> > http://dominion.thruhere.net/koen/OE/for-stable/
> >
> > What's the proper procedure for getting those in?
>
> And please answer in the form of a link to a wiki page :)

http://wiki.openembedded.net/index.php/Stable contains first 
informations.

Regards, 
-- 
JID:      hrw@jabber.org
Website:  http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz

Re: [STABLE] branch created: stable/2009

[Koen Kooi]

On 31-03-09 21:49, Marcin Juszkiewicz wrote:
> Dnia wtorek, 31 marca 2009 o 19:17:27 Tom Rini napisał(a):
>>> Great!
>>>
>>> I already have a few patches queued up for the stablebranch:
>>>
>>> http://dominion.thruhere.net/koen/OE/for-stable/
>>>
>>> What's the proper procedure for getting those in?
>>
>> And please answer in the form of a link to a wiki page :)
>
> http://wiki.openembedded.net/index.php/Stable contains first
> informations.

Thanks, to make getting the patches a bit easier I created a public 
patchwork bundle:

http://patchwork.openembedded.org/bundle/koen/koen-for-stable-20090331/

And if everyone uses the [STABLE] tag in mail subjects my patchwork 
filter will make it easier to pick them out.

regards,

Koen

[RFC] more streamlined review procedure, was: Re: [STABLE] branch created: stable/2009

[Koen Kooi]

On 31-03-09 21:49, Marcin Juszkiewicz wrote:
> Dnia wtorek, 31 marca 2009 o 19:17:27 Tom Rini napisał(a):
>>> Great!
>>>
>>> I already have a few patches queued up for the stablebranch:
>>>
>>> http://dominion.thruhere.net/koen/OE/for-stable/
>>>
>>> What's the proper procedure for getting those in?
>>
>> And please answer in the form of a link to a wiki page :)
>
> http://wiki.openembedded.net/index.php/Stable contains first
> informations.

I've found that there's a huge flaw in this setup:

Currently there are 8 people signed up as 'maintainers' for the stable 
branch, but only 2 (yes, two) have looked at some of the patches posted 
2 days ago. Another said he didn't want to look at patches for machines 
he didn't build for. No idea why the other 4 haven't responded, but if 
this continues then the stable branch can be closed down immediately.

Why? The current procedure requires an ACK from a stable 'maintainer' 
(not including yourself, of course), but you won't get an ACK or even a 
NACK.

This is annoying since I've had the first bugreports from users that 
were solved by patches that are 'under review'.

The previous stable branch tried to guarantee that you'd get at least a 
reaction on all your patches within 24 hours, so submitters knew what 
was happening. A 'reaction', not a 'review', so "will look at it next 
week" is perfectly well.

So my proposal:

* within 24 hours of posting at least 1 reaction from any stable 
'maintainer'
* No review within a week means automatic approval, commit must have 
"UNREVIEWED" marker to signify that.

Also:

I still have unreviewed patches in my for-stable bundle: 
http://patchwork.openembedded.org/bundle/koen/koen-for-stable-20090331/


regards,

Koen

Re: [RFC] more streamlined review procedure, was: Re: [STABLE] branch created: stable/2009

[Marco Cavallini]

Koen Kooi ha scritto:
> I've found that there's a huge flaw in this setup:
> 
> Currently there are 8 people signed up as 'maintainers' for the stable
> branch, but only 2 (yes, two) have looked at some of the patches posted
> 2 days ago. Another said he didn't want to look at patches for machines
> he didn't build for. No idea why the other 4 haven't responded, but if
> this continues then the stable branch can be closed down immediately.
> 
> Why? The current procedure requires an ACK from a stable 'maintainer'
> (not including yourself, of course), but you won't get an ACK or even a
> NACK.
> 
> This is annoying since I've had the first bugreports from users that
> were solved by patches that are 'under review'.
> 
> The previous stable branch tried to guarantee that you'd get at least a
> reaction on all your patches within 24 hours, so submitters knew what
> was happening. A 'reaction', not a 'review', so "will look at it next
> week" is perfectly well.
> 
> So my proposal:
> 
> * within 24 hours of posting at least 1 reaction from any stable
> 'maintainer'
> * No review within a week means automatic approval, commit must have
> "UNREVIEWED" marker to signify that.

Hi
I'm trying to understand the best way to proceed.
I think you shouldn't ACK a patch that you can't test.
Maybe I'm wrong or I'm misunderstanding the goal of a 'stable' branch?

/marco

Re: [RFC] more streamlined review procedure, was: Re: [STABLE] branch created: stable/2009

[Koen Kooi]

On 02-04-09 17:31, Marco Cavallini wrote:
> Koen Kooi ha scritto:
>> I've found that there's a huge flaw in this setup:
>>
>> Currently there are 8 people signed up as 'maintainers' for the stable
>> branch, but only 2 (yes, two) have looked at some of the patches posted
>> 2 days ago. Another said he didn't want to look at patches for machines
>> he didn't build for. No idea why the other 4 haven't responded, but if
>> this continues then the stable branch can be closed down immediately.
>>
>> Why? The current procedure requires an ACK from a stable 'maintainer'
>> (not including yourself, of course), but you won't get an ACK or even a
>> NACK.
>>
>> This is annoying since I've had the first bugreports from users that
>> were solved by patches that are 'under review'.
>>
>> The previous stable branch tried to guarantee that you'd get at least a
>> reaction on all your patches within 24 hours, so submitters knew what
>> was happening. A 'reaction', not a 'review', so "will look at it next
>> week" is perfectly well.
>>
>> So my proposal:
>>
>> * within 24 hours of posting at least 1 reaction from any stable
>> 'maintainer'
>> * No review within a week means automatic approval, commit must have
>> "UNREVIEWED" marker to signify that.
>
> Hi
> I'm trying to understand the best way to proceed.
> I think you shouldn't ACK a patch that you can't test.
> Maybe I'm wrong or I'm misunderstanding the goal of a 'stable' branch?

It seems you didn't read http://wiki.openembedded.net/index.php/Stable:

"It is important to note that we will test for building. Testing on 
target devices is left for users which can (and should) report bugs if 
something is not working. More about reporting bugs later in that text."

There's no "can't" when it comes for testing for the stable branch.

regards,

Koen

PS: I do think we should test on a target device whenever possible

Re: [STABLE] branch created: stable/2009

[Robert Schuster]

[-- Attachment #1: Type: text/plain, Size: 1411 bytes --]

Hi,
is it possible/allowed to port things into the stable which IMO make its
maintenance easier? E.g. I've just cleaned up the llvm recipes and made
them side by side installable (multiple version). The stable branch
would be more of a burden when these things differ so much from each other.

Apart from that llvm is not used from anything but the llvm-gcc recipes
which I saw and which look not inspiring confidence.

Last but not least I was going to base the Jalimo recipes on that and
then finally move them into OE.

Regards
Robert

Marcin Juszkiewicz schrieb:
> Hi
> 
> I would like to announce creation of new branch: stable/2009 in 
> OpenEmbedded repository. It was branched few days ago, build tested and 
> got some bugfixes which were added to .dev tree (and back to 
> stable/2009).
> 
> What was in build test? Distro "angstrom-2008.1", machines: qemuarm 
> (armv5te), at91sam9263ek (armv5te), beagleboard (armv7a), qemux86 (i586) 
> and vortex86sx (486sx). I tested building of next images: console, 
> console-base, x11, beagleboard-demo for each of those machines. So far 
> only glibc builds were tested, uclibc one is queued.
> 
> I also imported BitBake 1.8.12 (last release) into "stable/2009" branch 
> as this is minimal stable version which works with this repository.
> 
> Feel free to clone and do tests. I will write more later.
> 
> Regards, 



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 268 bytes --]

Re: [STABLE] branch created: stable/2009

[Marcin Juszkiewicz]

Dnia wtorek, 31 marca 2009 o 22:02:56 Robert Schuster napisał(a):
> is it possible/allowed to port things into the stable which IMO make
> its maintenance easier? E.g. I've just cleaned up the llvm recipes
> and made them side by side installable (multiple version). The stable
> branch would be more of a burden when these things differ so much
> from each other.

Read http://wiki.openembedded.net/index.php/Stable#Dealing_with_changes 
and provide patches.

> Apart from that llvm is not used from anything but the llvm-gcc
> recipes which I saw and which look not inspiring confidence.

> Last but not least I was going to base the Jalimo recipes on that and
> then finally move them into OE.

As long as your changes will also land in .dev (in same of different 
versions due to differences between .dev/stable) it will be ok.

Regards, 
-- 
JID:      hrw@jabber.org
Website:  http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz

Re: [STABLE] branch created: stable/2009

[Marco Cavallini]

Marcin Juszkiewicz ha scritto:
> Hi
> 
> I would like to announce creation of new branch: stable/2009 in 
> OpenEmbedded repository. It was branched few days ago, build tested and 
> got some bugfixes which were added to .dev tree (and back to 
> stable/2009).
> 
> What was in build test? Distro "angstrom-2008.1", machines: qemuarm 
> (armv5te), at91sam9263ek (armv5te), beagleboard (armv7a), qemux86 (i586) 
> and vortex86sx (486sx). I tested building of next images: console, 
> console-base, x11, beagleboard-demo for each of those machines. So far 
> only glibc builds were tested, uclibc one is queued.
> 
> I also imported BitBake 1.8.12 (last release) into "stable/2009" branch 
> as this is minimal stable version which works with this repository.
> 
> Feel free to clone and do tests. I will write more later.
> 
> Regards, 

What command do you use to get it ?

/marco

Re: [STABLE] branch created: stable/2009

[Esben Haabendal]

On Tue, Mar 31, 2009 at 6:37 PM, Marcin Juszkiewicz
<marcin@juszkiewicz.com.pl> wrote:
>
> I also imported BitBake 1.8.12 (last release) into "stable/2009" branch
> as this is minimal stable version which works with this repository.

Is this the new strategy for handling depedency on specific versions of BitBake?

/Esben

-- 
Esben Haabendal, Senior Software Consultant
DoréDevelopment ApS, Ved Stranden 1, 9560 Hadsund, DK-Denmark
Phone: +45 51 92 53 93, E-mail: eha@doredevelopment.dk
WWW: http://www.doredevelopment.dk

Re: [STABLE] branch created: stable/2009

[Marcin Juszkiewicz]

Dnia czwartek, 2 kwietnia 2009 o 12:52:51 Esben Haabendal napisał(a):
> On Tue, Mar 31, 2009 at 6:37 PM, Marcin Juszkiewicz
>
> <marcin@juszkiewicz.com.pl> wrote:
> > I also imported BitBake 1.8.12 (last release) into "stable/2009"
> > branch as this is minimal stable version which works with this
> > repository.
>
> Is this the new strategy for handling depedency on specific versions
> of BitBake?

No, but we require minimal versions of BitBake to be used with OE 
metadata so I prefer to have version which is granted to work. Imagine 
situation when OE.dev will require BitBake 1.8.22 which will not work 
with 'stable/2009' branch.

Regards, 
-- 
JID:      hrw@jabber.org
Website:  http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz

Re: [STABLE] branch created: stable/2009

[Esben Haabendal]

On Thu, Apr 2, 2009 at 1:25 PM, Marcin Juszkiewicz
<marcin@juszkiewicz.com.pl> wrote:
> Dnia czwartek, 2 kwietnia 2009 o 12:52:51 Esben Haabendal napisał(a):
>> On Tue, Mar 31, 2009 at 6:37 PM, Marcin Juszkiewicz
>>
>> <marcin@juszkiewicz.com.pl> wrote:
>> > I also imported BitBake 1.8.12 (last release) into "stable/2009"
>> > branch as this is minimal stable version which works with this
>> > repository.
>>
>> Is this the new strategy for handling depedency on specific versions
>> of BitBake?
>
> No, but we require minimal versions of BitBake to be used with OE
> metadata so I prefer to have version which is granted to work. Imagine
> situation when OE.dev will require BitBake 1.8.22 which will not work
> with 'stable/2009' branch.

I fully understand the situation, but I am just wondering if we shouldn't be
paying more attention to this problem.

For the stable branches I need to maintain, I have had to make a solution
control which version of bitbake is used with each branch. Maybe it would
be worth to try work out a more general version of such a solutin.

But, I think it would be great if we "encode" into the OpenEmbedded
metadata which version of bitbake is recommended, and maybe even
which versions work and which is not expected to work. It could be
something as simple as adding a ".bitbake" to the top of the
metadata which contains a single line in the style of
"tags/bitbake-1.8.10", which would then make it possible for
people to know which version to use, and even automate
this in their setups.

We could then also make bitbake aware of this file, and then print
out a warning if a "wrong" version is used.

I am sure I am not the only one who have had to debug strange
problems before finding out that it would just be easier to go back
to an older bitbake version...

It is not that it is a big problem to import BitBake into OpenEmbedded,
although it generally is not the best solution to many things, as it
makes it a bit painful to propagate eventual critical fixes to all
the places where it is needed, but I just think that this is an
important issue that is worth to find a proper solution for.

Regards,
-- 
Esben Haabendal, Senior Software Consultant
DoréDevelopment ApS, Ved Stranden 1, 9560 Hadsund, DK-Denmark
Phone: +45 51 92 53 93, E-mail: eha@doredevelopment.dk
WWW: http://www.doredevelopment.dk

Re: [STABLE] branch created: stable/2009

[Marcin Juszkiewicz]

Dnia piątek, 3 kwietnia 2009 o 18:18:49 Esben Haabendal napisał(a):
> > No, but we require minimal versions of BitBake to be used with OE
> > metadata so I prefer to have version which is granted to work.
> > Imagine situation when OE.dev will require BitBake 1.8.22 which
> > will not work with 'stable/2009' branch.

> I fully understand the situation, but I am just wondering if we
> shouldn't be paying more attention to this problem.
>
> For the stable branches I need to maintain, I have had to make a
> solution control which version of bitbake is used with each branch.
> Maybe it would be worth to try work out a more general version of
> such a solutin.
>
> But, I think it would be great if we "encode" into the OpenEmbedded
> metadata which version of bitbake is recommended, and maybe even
> which versions work and which is not expected to work. It could be
> something as simple as adding a ".bitbake" to the top of the
> metadata which contains a single line in the style of
> "tags/bitbake-1.8.10", which would then make it possible for
> people to know which version to use, and even automate
> this in their setups.
>
> We could then also make bitbake aware of this file, and then print
> out a warning if a "wrong" version is used.
>
> I am sure I am not the only one who have had to debug strange
> problems before finding out that it would just be easier to go back
> to an older bitbake version...
>
> It is not that it is a big problem to import BitBake into
> OpenEmbedded, although it generally is not the best solution to many
> things, as it makes it a bit painful to propagate eventual critical
> fixes to all the places where it is needed, but I just think that
> this is an important issue that is worth to find a proper solution
> for.

sanity.conf sets one variable: BB_MIN_VERSION = "1.8.12" and this is 
checked on start of each build. We bump that only when BitBake is 
released and if change is really required.


Regards, 
-- 
JID:      hrw@jabber.org
Website:  http://marcin.juszkiewicz.com.pl/
LinkedIn: http://www.linkedin.com/in/marcinjuszkiewicz

Re: [STABLE] branch created: stable/2009

[Esben Haabendal]

2009/4/3 Marcin Juszkiewicz <marcin@juszkiewicz.com.pl>:

> sanity.conf sets one variable: BB_MIN_VERSION = "1.8.12" and this is
> checked on start of each build. We bump that only when BitBake is
> released and if change is really required.

Ok.

Would it be a sane solution to them simply use the BitBake version specified
in BB_MIN_VERSION, or do the "if change is really required" mean that
it sometimes
makes good sense to use a newer version?

If so, it would be nice to have something like BB_RECOMMENDED_VERSION

Also, is the BB_MIN_VERSION defined so it will/must always be a tagged
version of bitbake?

/Esben