[Buildroot] Write Access to buildroot?

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

What happened to this?
Can't even generate a branch right now...
BR
Ulf Samuelsson

[Buildroot] Write Access to buildroot?

[Peter Korsgaard]

>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:

 Ulf> What happened to this?
 Ulf> Can't even generate a branch right now...

That's on purpose. See:

http://lists.busybox.net/pipermail/buildroot/2009-March/026875.html

and

http://buildroot.net/git.html

-- 
Bye, Peter Korsgaard

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

Peter Korsgaard skrev:
>>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:
>>>>>>             
>
>  Ulf> What happened to this?
>  Ulf> Can't even generate a branch right now...
>
> That's on purpose. See:
>
> http://lists.busybox.net/pipermail/buildroot/2009-March/026875.html
>
> and
>
> http://buildroot.net/git.html
>
>   
OK,
Have not been active for a while and have about 3000 unread emails.
Seems to me that there a multiple people with write access 
from the git commit mails, How come?

Personally, I think it would make sense to have
a development git with personal branches with write access.

This would complement any personal git's and would allow
easier testing with other people's patches.

The main tree, used for the release
could then pull changes from both the development tree
and the personal tree's.

Is there a list of maintainers?
The target/device/Atmel tree will need to be updated due to the SAM9G10/G45/M10/M11.
The target/linux tree is also in need of maintenance.


BR
Ulf Samuelsson

[Buildroot] Write Access to buildroot?

[Daniel Mack]

On Fri, Jul 24, 2009 at 10:43:06AM +0200, Ulf Samuelsson wrote:
> > http://lists.busybox.net/pipermail/buildroot/2009-March/026875.html
> >
> > and
> >
> > http://buildroot.net/git.html
> >
> >   
> OK,
> Have not been active for a while and have about 3000 unread emails.
> Seems to me that there a multiple people with write access 
> from the git commit mails, How come?

That's just the git post-commit hook that takes the original author as
sender address of the email. That always happens when Peter picks a
patch and commits it to his tree.

> Personally, I think it would make sense to have
> a development git with personal branches with write access.
> 
> This would complement any personal git's and would allow
> easier testing with other people's patches.

The thing about git and other distributed SCMs is: everbody has write
access. To his local copy. You can than push your tree to an publically
accesible external tree and share your personal patches with others.
Or do the same thing by sending a thread of patches, if you don't want
to share your patches among others before they're ready for the master
branch.

> The main tree, used for the release
> could then pull changes from both the development tree
> and the personal tree's.

This is basicaly exactly the same thing than having your own repository,
with the difference that the admins of ublibc.org won't have to fiddle
with user permissions.

> Is there a list of maintainers?
> The target/device/Atmel tree will need to be updated due to the SAM9G10/G45/M10/M11.
> The target/linux tree is also in need of maintenance.

That's exactly why you want to have distributed repositories. You simply
don't have to care about all that. Once you tested the patches in your
tree for long enough, simply send in a pull request, and they will end
up in the master branch.

There a multiple free git hosting services out there, for example
github.com.

Hope that helps :)

Daniel

[Buildroot] Write Access to buildroot?

[Thomas Petazzoni]

Le Fri, 24 Jul 2009 10:56:03 +0200,
Daniel Mack <daniel@caiaq.de> a ?crit :

> There a multiple free git hosting services out there, for example
> github.com.

As explained in
http://lists.busybox.net/pipermail/buildroot/2009-March/026875.html,
Buildroot developers also have a public git repository on the uclibc
server. For example, my own git tree is visible at
http://git.buildroot.net/~tpetazzoni/git/buildroot/.

Sincerly,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers and embedded Linux development,
consulting, training and support.
http://free-electrons.com

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

> On Fri, Jul 24, 2009 at 10:43:06AM +0200, Ulf Samuelsson wrote:
>> > http://lists.busybox.net/pipermail/buildroot/2009-March/026875.html
>> >
>> > and
>> >
>> > http://buildroot.net/git.html
>> >
>> >
>> OK,
>> Have not been active for a while and have about 3000 unread emails.
>> Seems to me that there a multiple people with write access
>> from the git commit mails, How come?
>
> That's just the git post-commit hook that takes the original author as
> sender address of the email. That always happens when Peter picks a
> patch and commits it to his tree.
>
>> Personally, I think it would make sense to have
>> a development git with personal branches with write access.
>>
>> This would complement any personal git's and would allow
>> easier testing with other people's patches.
>
> The thing about git and other distributed SCMs is: everbody has write
> access. To his local copy. You can than push your tree to an publically
> accesible external tree and share your personal patches with others.

That is the problem, buildroot does not provide a "publically accesible 
external tree"
for developers.
Why rely on github.com mentioned below which people won't find on their own?

If the development tree is within buildroot, then people are more likely to 
test it
or contribute.


> Or do the same thing by sending a thread of patches, if you don't want
> to share your patches among others before they're ready for the master
> branch.
>
>> The main tree, used for the release
>> could then pull changes from both the development tree
>> and the personal tree's.
>
> This is basicaly exactly the same thing than having your own repository,
> with the difference that the admins of ublibc.org won't have to fiddle
> with user permissions.

The difference is that the development tree is shared.

>
>> Is there a list of maintainers?
>> The target/device/Atmel tree will need to be updated due to the 
>> SAM9G10/G45/M10/M11.
>> The target/linux tree is also in need of maintenance.
>
> That's exactly why you want to have distributed repositories. You simply
> don't have to care about all that. Once you tested the patches in your
> tree for long enough, simply send in a pull request, and they will end
> up in the master branch.
>
> There a multiple free git hosting services out there, for example
> github.com.

Thanks for the link.


>
> Hope that helps :)
>
> Daniel
>


Best Regards
Ulf Samuelsson                ulf at atmel.com
Atmel Nordic AB
Mail:  Box 2033, 174 02 Sundbyberg, Sweden
Visit:  Kavalleriv?gen 24, 174 58 Sundbyberg, Sweden
Phone +46 (8) 441 54 22     Fax +46 (8) 441 54 29
GSM    +46 (706) 22 44 57

[Buildroot] Write Access to buildroot?

[Thomas Petazzoni]

Le Fri, 24 Jul 2009 11:19:36 +0200,
"Ulf Samuelsson" <ulf.samuelsson@atmel.com> a ?crit :

> That is the problem, buildroot does not provide a "publically
> accesible external tree"
> for developers.

As explained on
http://lists.busybox.net/pipermail/buildroot/2009-March/026875.html,
there are publically accessible external trees provided by Buildroot
for developers.

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers and embedded Linux development,
consulting, training and support.
http://free-electrons.com

[Buildroot] Write Access to buildroot?

[Thomas Petazzoni]

[ Keeping the list in Cc: since these things should probably be
  discussed publicly. ]

Le Fri, 24 Jul 2009 13:37:01 +0200,
Ulf Samuelsson <ulf.samuelsson@atmel.com> a ?crit :

> With read access only, allowing you to make your own tree from the
> main tree.
> If you want to share your stuff, before committing to the main tree,
> you have to do it outside the buildroot world.
> Am I missing something?

Yes you're missing something.

Look at http://git.buildroot.net/

You have a link to the official Buildroot Git tree, browsable at
http://git.buildroot.net/buildroot/. This Git tree is read-only, except
for Peter.

But at the bottom of the page, you have ? User repositories ?. These
are read/write for their owner. For example, mine is
http://git.buildroot.net/~tpetazzoni/git/buildroot/. In my own
repository, I have a master branch which is just a clone of the main
Buildroot repository. And all the other branches (presently
directfb-bump, ext-toolchain-target-path, external-toolchain,
kconfig-update and xorg) contain my *own* work, ready for submission to
Peter. And some of this work has actually already been integrated in
mainline (branches directfb-bump, ext-toolchain-target-path and
external-toolchain).

The link
http://lists.busybox.net/pipermail/buildroot/2009-March/026875.html
that I've already posted explains how to use these user repositories,
that *are read/write.

Does that clarify the situation ?

Sincerly,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers and embedded Linux development,
consulting, training and support.
http://free-electrons.com

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

Thomas Petazzoni skrev:
> [ Keeping the list in Cc: since these things should probably be
>   discussed publicly. ]
>
> Le Fri, 24 Jul 2009 13:37:01 +0200,
> Ulf Samuelsson <ulf.samuelsson@atmel.com> a ?crit :
>
>   
>> With read access only, allowing you to make your own tree from the
>> main tree.
>> If you want to share your stuff, before committing to the main tree,
>> you have to do it outside the buildroot world.
>> Am I missing something?
>>     
>
> Yes you're missing something.
>
> Look at http://git.buildroot.net/
>
> You have a link to the official Buildroot Git tree, browsable at
> http://git.buildroot.net/buildroot/. This Git tree is read-only, except
> for Peter.
>
> But at the bottom of the page, you have ? User repositories ?. These
> are read/write for their owner. For example, mine is
> http://git.buildroot.net/~tpetazzoni/git/buildroot/. In my own
> repository, I have a master branch which is just a clone of the main
> Buildroot repository. And all the other branches (presently
> directfb-bump, ext-toolchain-target-path, external-toolchain,
> kconfig-update and xorg) contain my *own* work, ready for submission to
> Peter. And some of this work has actually already been integrated in
> mainline (branches directfb-bump, ext-toolchain-target-path and
> external-toolchain).
>
> The link
> http://lists.busybox.net/pipermail/buildroot/2009-March/026875.html
> that I've already posted explains how to use these user repositories,
> that *are read/write.
>
> Does that clarify the situation ?
>
>   
And what I am missing, is a way to have several people commit to the
same tree.
I.E: A development tree.

> Sincerly,
>
> Thomas
>   
BR
Ulf Samuelsson

[Buildroot] Write Access to buildroot?

[Thomas Petazzoni]

Le Fri, 24 Jul 2009 17:16:30 +0200,
Ulf Samuelsson <ulf.samuelsson@atmel.com> a ?crit :

> And what I am missing, is a way to have several people commit to the
> same tree.
> I.E: A development tree.

This is not how development works with Git. Everyone works in its
own working tree, and asks the project maintainer to pull the working
tree when things are ready. Things gets integrated in the project
maintainer tree, which everyone can update from to get the changes from
the other developers.

In Buildroot, there's no ? development tree ? in the current
development model. The same tree (the official way) is sometimes a
development tree (before -rc1 is released), sometimes a stabilization
tree (between the release of -rc1 and the release of the stable
version).

Sincerly,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers and embedded Linux development,
consulting, training and support.
http://free-electrons.com

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

Thomas Petazzoni skrev:
> Le Fri, 24 Jul 2009 17:16:30 +0200,
> Ulf Samuelsson <ulf.samuelsson@atmel.com> a ?crit :
>
>   
>> And what I am missing, is a way to have several people commit to the
>> same tree.
>> I.E: A development tree.
>>     
>
> This is not how development works with Git. Everyone works in its
> own working tree, and asks the project maintainer to pull the working
> tree when things are ready. Things gets integrated in the project
> maintainer tree, which everyone can update from to get the changes from
> the other developers.
>   
In the openembedded project all the developers have
write access to the development tree and can create
their own branches within the tree.

This makes it a lot easier to test other peoples patches.

If you look at the statistics for the user repositories,
you see that there are 5 people with user repositories,
including mine, and last updates were from

correa:          3 months ago,
wberrier:       2 months ago,
tpetazzoni:   9 days ago
ulf:                 33 hours ago
jacmet:         -

This does not indicate a high level of popularity to me.
Maybe people are keeping trees in other locations, but
that means that they are hidden from most users.

In reality, the way Buildroot works today seems to be that
people are sending patches to the mailing-list.

It is fairly easy to set up common git trees,
by creating a "common" user with a git subdirectory
and then have a symbolic link in each user directory
to this git directory with group write permissions.

> In Buildroot, there's no ? development tree ? in the current
> development model. The same tree (the official way) is sometimes a
> development tree (before -rc1 is released), sometimes a stabilization
> tree (between the release of -rc1 and the release of the stable
> version).
>
>   
The suggestion does not change this.
> Sincerly,
>
> Thomas
>   
BR
Ulf Samuelsson

[Buildroot] Write Access to buildroot?

[Peter Korsgaard]

>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:

Hi,

 Ulf> In the openembedded project all the developers have
 Ulf> write access to the development tree and can create
 Ulf> their own branches within the tree.

 Ulf> This makes it a lot easier to test other peoples patches.

I don't agree, typing git add remove <foo> is trivial. That's the
glory of the distributed nature of git.

 Ulf> If you look at the statistics for the user repositories,
 Ulf> you see that there are 5 people with user repositories,
 Ulf> including mine, and last updates were from

 Ulf> correa:          3 months ago,
 Ulf> wberrier:       2 months ago,
 Ulf> tpetazzoni:   9 days ago
 Ulf> ulf:                 33 hours ago
 Ulf> jacmet:         -

which corresponds pretty nicely with the list of active developers:

git shortlog -s -n 2009.05..
    73  Peter Korsgaard
    31  Thomas Petazzoni
     9  Maxim Grigoriev
     8  Will Newton
     7  Sven Neumann
     5  Gustavo Zacarias
     4  Nigel Kukard
     2  Bernhard Reutner-Fischer
     2  Daniel Mack
     2  Ulf Samuelsson
     1  Alex Dobrynin
     1  Evan Zelkowitz
     1  H Hartley Sweeten
     1  Lionel Landwerlin
     1  Markus Heidelberg
     1  Pwalters
     1  laurent laffont

Bernhard has his repo on repo.or.cz for historical reasons.

 Ulf> This does not indicate a high level of popularity to me.
 Ulf> Maybe people are keeping trees in other locations, but
 Ulf> that means that they are hidden from most users.

No, it indicates that buildroot is a small project with limited amount
of active developers (and luckily a lot of outside contributers).

 Ulf> In reality, the way Buildroot works today seems to be that
 Ulf> people are sending patches to the mailing-list.

And that's exactly what they should do. We *WANT* to see the patches
on the mailing list so they can get reviewed.

We also don't want to require outsiders to learn git in order to
contribute patches.

-- 
Bye, Peter Korsgaard

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

Peter Korsgaard skrev:
>>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:
>>>>>>             
>
> Hi,
>
>  Ulf> In the openembedded project all the developers have
>  Ulf> write access to the development tree and can create
>  Ulf> their own branches within the tree.
>
>  Ulf> This makes it a lot easier to test other peoples patches.
>
> I don't agree, typing git add remove <foo> is trivial. That's the
> glory of the distributed nature of git.
>
>  Ulf> If you look at the statistics for the user repositories,
>  Ulf> you see that there are 5 people with user repositories,
>  Ulf> including mine, and last updates were from
>
>  Ulf> correa:          3 months ago,
>  Ulf> wberrier:       2 months ago,
>  Ulf> tpetazzoni:   9 days ago
>  Ulf> ulf:                 33 hours ago
>  Ulf> jacmet:         -
>
> which corresponds pretty nicely with the list of active developers:
>
> git shortlog -s -n 2009.05..
>     73  Peter Korsgaard
>     31  Thomas Petazzoni
>      9  Maxim Grigoriev
>      8  Will Newton
>      7  Sven Neumann
>      5  Gustavo Zacarias
>      4  Nigel Kukard
>      2  Bernhard Reutner-Fischer
>      2  Daniel Mack
>      2  Ulf Samuelsson
>      1  Alex Dobrynin
>      1  Evan Zelkowitz
>      1  H Hartley Sweeten
>      1  Lionel Landwerlin
>      1  Markus Heidelberg
>      1  Pwalters
>      1  laurent laffont
>
> Bernhard has his repo on repo.or.cz for historical reasons.
>
>  Ulf> This does not indicate a high level of popularity to me.
>  Ulf> Maybe people are keeping trees in other locations, but
>  Ulf> that means that they are hidden from most users.
>
> No, it indicates that buildroot is a small project with limited amount
> of active developers (and luckily a lot of outside contributers).
>
>  Ulf> In reality, the way Buildroot works today seems to be that
>  Ulf> people are sending patches to the mailing-list.
>
> And that's exactly what they should do. We *WANT* to see the patches
> on the mailing list so they can get reviewed.
>   
I do not see a contracdiction here.
You can have multiple persons work on one thing, and
once it is ready, the patches can be automatically
generated and sent to the mailing list.

What is the drawback of letting people work together
by allowing them to share a git.

People can share a git anyway, by using another git server,
but that makes it less visible to Buildroot users, so I would
like to avoid that.

> We also don't want to require outsiders to learn git in order to
> contribute patches.
>   
It it very different to require people to learn git, from
blocking people that wants to use git, to work together.

BR
Ulf Samuelsson

[Buildroot] Write Access to buildroot?

[Peter Korsgaard]

>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:

Hi,

 >> We also don't want to require outsiders to learn git in order to
 >> contribute patches.
 >> 
 Ulf> It it very different to require people to learn git, from
 Ulf> blocking people that wants to use git, to work together.

Blocking? Noone is blocking anyone from anything. That particular work
flow is not used for the official tree, but you can do what you want
to with your own trees.

Don't confuse working together with commit access. The Linux kernel
gets contributions from ~1000 people for each release, and yet Linus
is the only one having commit access to the official tree.

-- 
Bye, Peter Korsgaard

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

Peter Korsgaard skrev:
>>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:
> 
> Hi,
> 
>  >> We also don't want to require outsiders to learn git in order to
>  >> contribute patches.
>  >> 
>  Ulf> It it very different to require people to learn git, from
>  Ulf> blocking people that wants to use git, to work together.
> 
> Blocking? Noone is blocking anyone from anything. That particular work
> flow is not used for the official tree, but you can do what you want
> to with your own trees.
> 

I have requested a capability where multiple people have
write capability to a tree which can then be pulled
to the official tree.

It is possible to do this on another git server,
but that means that it is less visible to other Buildroot users.

It is simple to implement.
Create the "common" user, and let developers have
write access to "~common/git", possibly using a symbolic link.

> Don't confuse working together with commit access. The Linux kernel
> gets contributions from ~1000 people for each release, and yet Linus
> is the only one having commit access to the official tree.
> 

That comments is irrelevant because I am not asking for commit
access to the official tree.

BR
Ulf Samuelsson

[Buildroot] Write Access to buildroot?

[Peter Korsgaard]

>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:

Hi,

 Ulf> I have requested a capability where multiple people have
 Ulf> write capability to a tree which can then be pulled
 Ulf> to the official tree.

 Ulf> It is possible to do this on another git server,
 Ulf> but that means that it is less visible to other Buildroot users.

 Ulf> It is simple to implement.
 Ulf> Create the "common" user, and let developers have
 Ulf> write access to "~common/git", possibly using a symbolic link.

Then do so. You have shell access to the buildroot.net machine, so you
can make ~ulf/git/common world writable.

-- 
Bye, Peter Korsgaard

[Buildroot] Write Access to buildroot?

[Marc Gauthier]

For what it's worth, we have a shared git tree on linux-xtensa.org
for Xtensa arch development of buildroot (we've had it there for
many months before the buildroot project itself switched over to git).

We had to be careful with file ownership/protection (and maybe umask)
to make sure it works smoothly, but once setup, it works quite well.
Everybody has their own userid, no shared user or anything.
We always use the git+ssh://... method to access the tree if we want
to be able to write to it, else anyone can access it read-only using
the usual git://... method -- if you're curious, see:

   http://wiki.linux-xtensa.org/index.php/Repository_Access

One also has to be careful to only push things to this tree.
No playing directly with the tree (such as rebase or otherwise
pointing HEAD backwards in time and possibly committing new changes
from there) which plays havoc with others' clones of the tree.
As long as it's only accessed remotely, this is pretty easy.

In case you notice it out of sync... we've only merged this tree with
the official buildroot tree once in a long while -- were waiting for
the Xtensa arch port to get submitted (which it just has, yeah! :-) and
to setup things to make it easier to catch merges that break things.

-Marc


Ulf Samuelsson wrote:
> Peter Korsgaard skrev:
> >>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:
> >
> > Hi,
> >
> >  >> We also don't want to require outsiders to learn git in order to
> >  >> contribute patches.
> >  >>
> >  Ulf> It it very different to require people to learn git, from
> >  Ulf> blocking people that wants to use git, to work together.
> >
> > Blocking? Noone is blocking anyone from anything. That
> particular work
> > flow is not used for the official tree, but you can do what you want
> > to with your own trees.
> >
>
> I have requested a capability where multiple people have
> write capability to a tree which can then be pulled
> to the official tree.
>
> It is possible to do this on another git server,
> but that means that it is less visible to other Buildroot users.
>
> It is simple to implement.
> Create the "common" user, and let developers have
> write access to "~common/git", possibly using a symbolic link.
>
> > Don't confuse working together with commit access. The Linux kernel
> > gets contributions from ~1000 people for each release, and yet Linus
> > is the only one having commit access to the official tree.
> >
>
> That comments is irrelevant because I am not asking for commit
> access to the official tree.
>
> BR
> Ulf Samuelsson
> _______________________________________________
> buildroot mailing list
> buildroot at busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot
>

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

>

> For what it's worth, we have a shared git tree on linux-xtensa.org
> for Xtensa arch development of buildroot (we've had it there for
> many months before the buildroot project itself switched over to git).

> We had to be careful with file ownership/protection (and maybe umask)
> to make sure it works smoothly, but once setup, it works quite well.
> Everybody has their own userid, no shared user or anything.

Thanks, it is the model used by Openembedded as well.

I do not specifically need a shared user,
I just want the http://git.uclibc.org/ to report

~common/<myproject>

instead of

~/ulf/git/common/<myproject>.

since the former better illustrates how I'd like to work.

It seems easier to just create a new user though.

> We always use the git+ssh://... method to access the tree if we want
> to be able to write to it, else anyone can access it read-only using
> the usual git://... method -- if you're curious, see:

>    http://wiki.linux-xtensa.org/index.php/Repository_Access

> One also has to be careful to only push things to this tree.
> No playing directly with the tree (such as rebase or otherwise
> pointing HEAD backwards in time and possibly committing new changes
> from there) which plays havoc with others' clones of the tree.
> As long as it's only accessed remotely, this is pretty easy.

> In case you notice it out of sync... we've only merged this tree with
> the official buildroot tree once in a long while -- were waiting for
> the Xtensa arch port to get submitted (which it just has, yeah! :-) and
> to setup things to make it easier to catch merges that break things.

> -Marc



Best Regards
Ulf Samuelsson                ulf at atmel.com
Atmel Nordic AB
Mail:  Box 2033, 174 02 Sundbyberg, Sweden
Visit:  Kavalleriv?gen 24, 174 58 Sundbyberg, Sweden
Phone +46 (8) 441 54 22     Fax +46 (8) 441 54 29
GSM    +46 (706) 22 44 57

[Buildroot] Write Access to buildroot?

[Thomas Petazzoni]

Le Mon, 27 Jul 2009 08:40:13 +0200,
"Ulf Samuelsson" <ulf.samuelsson@atmel.com> a ?crit :

> I just want the http://git.uclibc.org/ to report
> 
> ~common/<myproject>
> 
> instead of
> 
> ~/ulf/git/common/<myproject>.
> 
> since the former better illustrates how I'd like to work.

I really don't see the difference between the two approaches, and what
are the advantages of the approach you're proposing. In terms of
visibility, it is *exactly* the same thing.

Honestly, what I understand from this discussion is that you're saying
?I want this to be *this* way and no other way ?, but you're not giving
any strong and convicing arguments of why the current way is bad and
should be changed.

The way we do it in Buildroot is similar to dozens or hundreds of
projects out there. It may be different to dozens or hundreds other
projects. Participating to a community means that making concessions is
necessary to progress and work together.

Sincerly,

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers and embedded Linux development,
consulting, training and support.
http://free-electrons.com

[Buildroot] Write Access to buildroot?

[Ulf Samuelsson]

>

> For what it's worth, we have a shared git tree on linux-xtensa.org
> for Xtensa arch development of buildroot (we've had it there for
> many months before the buildroot project itself switched over to git).

> We had to be careful with file ownership/protection (and maybe umask)
> to make sure it works smoothly, but once setup, it works quite well.
> Everybody has their own userid, no shared user or anything.

Thanks, it is the model used by Openembedded as well.

I do not specifically need a shared user,
I just want the http://git.uclibc.org/ to report

~common/<myproject>

instead of

~/ulf/git/common/<myproject>.

since the former better illustrates how I'd like to work.

It seems easier to just create a new user though.

> We always use the git+ssh://... method to access the tree if we want
> to be able to write to it, else anyone can access it read-only using
> the usual git://... method -- if you're curious, see:

>    http://wiki.linux-xtensa.org/index.php/Repository_Access

> One also has to be careful to only push things to this tree.
> No playing directly with the tree (such as rebase or otherwise
> pointing HEAD backwards in time and possibly committing new changes
> from there) which plays havoc with others' clones of the tree.
> As long as it's only accessed remotely, this is pretty easy.

> In case you notice it out of sync... we've only merged this tree with
> the official buildroot tree once in a long while -- were waiting for
> the Xtensa arch port to get submitted (which it just has, yeah! :-) and
> to setup things to make it easier to catch merges that break things.

> -Marc



Best Regards
Ulf Samuelsson                ulf at atmel.com
Atmel Nordic AB
Mail:  Box 2033, 174 02 Sundbyberg, Sweden
Visit:  Kavalleriv?gen 24, 174 58 Sundbyberg, Sweden
Phone +46 (8) 441 54 22     Fax +46 (8) 441 54 29
GSM    +46 (706) 22 44 57

[Buildroot] Write Access to buildroot?

[Thomas Petazzoni]

Le Sun, 26 Jul 2009 22:21:00 +0200,
Ulf Samuelsson <ulf.samuelsson@atmel.com> a ?crit :

> It is possible to do this on another git server,
> but that means that it is less visible to other Buildroot users.

Since visibility seems to be your main complaint against the developer
Git trees, what kind of better visibility would you like than the one
already available at http://git.buildroot.net/ ?

All developer trees are directly visible, on the mainpage of the Git
page. Isn't that enough ?

Thomas
-- 
Thomas Petazzoni, Free Electrons
Kernel, drivers and embedded Linux development,
consulting, training and support.
http://free-electrons.com

[Buildroot] Write Access to buildroot?

[Peter Korsgaard]

>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:

Hi,

 Ulf> Personally, I think it would make sense to have
 Ulf> a development git with personal branches with write access.

You have the possibility to put your own development tree on the
server as explained by Thomas.

 Ulf> This would complement any personal git's and would allow
 Ulf> easier testing with other people's patches.

 Ulf> The main tree, used for the release
 Ulf> could then pull changes from both the development tree
 Ulf> and the personal tree's.

 Ulf> Is there a list of maintainers?

Not an explicit list, no.

 Ulf> The target/device/Atmel tree will need to be updated due to the
 Ulf> SAM9G10/G45/M10/M11.

You are probably in the best position to do this - Last I looked the
G45 wasn't publicly announced and no documentation available.

Patches are welcome.

 Ulf> The target/linux tree is also in need of maintenance.

What exactly do you mean? 2.6.31-rcX options?

-- 
Bye, Peter Korsgaard

[Buildroot] Write Access to buildroot?

[Peter Korsgaard]

>>>>> "Ulf" == Ulf Samuelsson <ulf.samuelsson@atmel.com> writes:

Hi,

 Ulf> The target/device/Atmel tree will need to be updated due to the
 Ulf> SAM9G10/G45/M10/M11.
 >> 
 >> You are probably in the best position to do this - Last I looked the
 >> G45 wasn't publicly announced and no documentation available.

 Ulf> I dont have a datasheet either, only design specs, but I do have
 Ulf> a board and there are internal patches available.

External as well. Nicolas Ferre posted 9G45 patches to
linux-arm-kernel back in June.

-- 
Bye, Peter Korsgaard