Re: SVN woes persist

Re: SVN woes persist

[Richard Purdie]

On Thu, 2010-10-07 at 17:15 -0600, Gary Thomas wrote:
> I updated to include these changes and gave it a try.  Here's what I did:
> 
>    % mkdir /work/new_build
>    % cd /work/poky_tree
>    % . poky-init-build-env /work/new_build
>      ... adjust DL_DIR, etc in conf/local.conf
>          SSTATE_MIRRORS is not defined yet
>    % bitbake some-image
> 
>    % mv sstate-cache sstate
>    % mv tmp tmp.old
>      ... adjust conf/local.conf to set SSTATE_MIRRORS like this
>          SSTATE_MIRRORS ?= "\
>          file://.* file:///work/new_build/sstate/"
> 
>    % bitbake some-image
> 
> First of all, it took an incredible length of time (maybe a minute on my fast machine)
> to even move past the resolve step.  Then it went off spewing lines about the state files
> (sadly I didn't save them).  It then proceeded to run through some 322 steps, telling me
> more about state files.  This took quite some time.  Then it seemed to just start the
> normal build all over, apparently without regard to any of the sstate info.
> 
> Did I miss something?  Is this how it should work?
> 
> Sorry to be such a pain over this, but I'd really like to see this part working.
> If there's something I can help with it, please let me know - I know my way around
> Python code, but I don't know all the internals of how you expect this to work.

Sorry not to get back to you sooner, I'm being pulled in a few different
directions and we're coming up to release.

In my own tests I've noticed a few issues and I've just pushed a number
of fixes into the master branch which should help in various areas. 

With regard to your specific test and problem, it shouldn't rebuild
everything. It turns out that I'd tested with http:// mirror urls and
not file:// ones which needed and extra piece of code to handle them
correctly.

Have another try with master and see if things are better this time!

Cheers,

Richard

Re: SVN woes persist

[Gary Thomas]

On 10/22/2010 07:50 AM, Richard Purdie wrote:
> On Thu, 2010-10-07 at 17:15 -0600, Gary Thomas wrote:
>> I updated to include these changes and gave it a try.  Here's what I did:
>>
>>     % mkdir /work/new_build
>>     % cd /work/poky_tree
>>     % . poky-init-build-env /work/new_build
>>       ... adjust DL_DIR, etc in conf/local.conf
>>           SSTATE_MIRRORS is not defined yet
>>     % bitbake some-image
>>
>>     % mv sstate-cache sstate
>>     % mv tmp tmp.old
>>       ... adjust conf/local.conf to set SSTATE_MIRRORS like this
>>           SSTATE_MIRRORS ?= "\
>>           file://.* file:///work/new_build/sstate/"
>>
>>     % bitbake some-image
>>
>> First of all, it took an incredible length of time (maybe a minute on my fast machine)
>> to even move past the resolve step.  Then it went off spewing lines about the state files
>> (sadly I didn't save them).  It then proceeded to run through some 322 steps, telling me
>> more about state files.  This took quite some time.  Then it seemed to just start the
>> normal build all over, apparently without regard to any of the sstate info.
>>
>> Did I miss something?  Is this how it should work?
>>
>> Sorry to be such a pain over this, but I'd really like to see this part working.
>> If there's something I can help with it, please let me know - I know my way around
>> Python code, but I don't know all the internals of how you expect this to work.
>
> Sorry not to get back to you sooner, I'm being pulled in a few different
> directions and we're coming up to release.
>
> In my own tests I've noticed a few issues and I've just pushed a number
> of fixes into the master branch which should help in various areas.
>
> With regard to your specific test and problem, it shouldn't rebuild
> everything. It turns out that I'd tested with http:// mirror urls and
> not file:// ones which needed and extra piece of code to handle them
> correctly.
>
> Have another try with master and see if things are better this time!

Will do.

Should this address the similar issue with GIT packages.  These seem
to be updating every time as well:
   git_github.com.wrpseudo.pseudo.git.tar.gz
   git_git.pokylinux.org.wr-kernel-tools.git.tar.gz
   git_git.pokylinux.org.prelink-cross.git.tar.gz

I'm still hoping for a way to have a static (no network needed) setup
soon.  I have to provide such in a training for some of my customers
in about a month :-)

Thanks

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

Re: SVN woes persist

[Richard Purdie]

On Fri, 2010-10-22 at 08:04 -0600, Gary Thomas wrote:
> Will do.
> 
> Should this address the similar issue with GIT packages.  These seem
> to be updating every time as well:
>    git_github.com.wrpseudo.pseudo.git.tar.gz
>    git_git.pokylinux.org.wr-kernel-tools.git.tar.gz
>    git_git.pokylinux.org.prelink-cross.git.tar.gz

There are changes in master which will stop everything except the
linux-wrs recipe doing that. The linux-wrs recipe uses "fullclone" mode
of the git fetcher and this has known issues which we will address soon.

If anything other than linux-wrs is refreshing on master other than that
recipe, let us know.

> I'm still hoping for a way to have a static (no network needed) setup
> soon.  I have to provide such in a training for some of my customers
> in about a month :-)

It is meant to work like that!

Cheers,

Richard

Re: SVN woes persist

[Gary Thomas]

On 10/22/2010 08:29 AM, Richard Purdie wrote:
> On Fri, 2010-10-22 at 08:04 -0600, Gary Thomas wrote:
>> Will do.
>>
>> Should this address the similar issue with GIT packages.  These seem
>> to be updating every time as well:
>>     git_github.com.wrpseudo.pseudo.git.tar.gz
>>     git_git.pokylinux.org.wr-kernel-tools.git.tar.gz
>>     git_git.pokylinux.org.prelink-cross.git.tar.gz
>
> There are changes in master which will stop everything except the
> linux-wrs recipe doing that. The linux-wrs recipe uses "fullclone" mode
> of the git fetcher and this has known issues which we will address soon.
>
> If anything other than linux-wrs is refreshing on master other than that
> recipe, let us know.

Looks good.  The one one that got updated today with a full build was pseudo
and that's because the pinned revision was changed since I last merged.

>> I'm still hoping for a way to have a static (no network needed) setup
>> soon.  I have to provide such in a training for some of my customers
>> in about a month :-)
>
> It is meant to work like that!

This part is working much better now, thanks.

Any progress on a way to hot-start things with the staged packages?
When last I tried it, it brought my system to its knees with no benefit
(see original message from 2010-10-07)

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------

Re: SVN woes persist

[Richard Purdie]

On Fri, 2010-10-22 at 11:55 -0600, Gary Thomas wrote:
> On 10/22/2010 08:29 AM, Richard Purdie wrote:
> > On Fri, 2010-10-22 at 08:04 -0600, Gary Thomas wrote:
> >> Will do.
> >>
> >> Should this address the similar issue with GIT packages.  These seem
> >> to be updating every time as well:
> >>     git_github.com.wrpseudo.pseudo.git.tar.gz
> >>     git_git.pokylinux.org.wr-kernel-tools.git.tar.gz
> >>     git_git.pokylinux.org.prelink-cross.git.tar.gz
> >
> > There are changes in master which will stop everything except the
> > linux-wrs recipe doing that. The linux-wrs recipe uses "fullclone" mode
> > of the git fetcher and this has known issues which we will address soon.
> >
> > If anything other than linux-wrs is refreshing on master other than that
> > recipe, let us know.
> 
> Looks good.  The one one that got updated today with a full build was pseudo
> and that's because the pinned revision was changed since I last merged.

Right, once its generated its tarball, it shouldn't touch the repo
again.

> >> I'm still hoping for a way to have a static (no network needed) setup
> >> soon.  I have to provide such in a training for some of my customers
> >> in about a month :-)
> >
> > It is meant to work like that!
> 
> This part is working much better now, thanks.
> 
> Any progress on a way to hot-start things with the staged packages?
> When last I tried it, it brought my system to its knees with no benefit
> (see original message from 2010-10-07)

I thought my last email made this clear, we've fixed issues, it works
better now.

Unfortunately there is an issue with the bits of the toolchain
overwriting each other so a build using the packages isn't entirely
working but you can for example generate images using them. All the
pieces are there and we just need to iron out a few bugs.

Cheers,

Richard