Eclipse Community Forums
Forum Search:

Search      Help    Register    Login    Home
Home » Eclipse Projects » Buckminster » [git-reader] Materializing components from different sub-paths of the same repository
[git-reader] Materializing components from different sub-paths of the same repository [message #654845] Thu, 17 February 2011 12:21 Go to next message
Daniel Weber is currently offline Daniel Weber
Messages: 51
Registered: July 2009
Member
Hi,

I am trying to set up a cquery + rmap for etrice. Since plugins are
located in different subdirectories in the repository, I created
separate searchPaths, like this:

<rm:searchPath name="etrice.examples">
<rm:provider componentTypes="osgi.bundle" readerType="git">
<rm:uri format="{0}/org.eclipse.etrice,examples/{1}">
<bc:propertyRef key="workspace.root" />
<bc:propertyRef key="buckminster.component" />
</rm:uri>
<rm:property key="git.remote.uri" value="${etrice.git.repo}"/>
</rm:provider>
</rm:searchPath>

<rm:searchPath name="etrice.modellib">
<rm:provider componentTypes="osgi.bundle" readerType="git">
<rm:uri format="{0}/org.eclipse.etrice,modellib/{1}">
<bc:propertyRef key="workspace.root" />
<bc:propertyRef key="buckminster.component" />
</rm:uri>
<rm:property key="git.remote.uri" value="${etrice.git.repo}"/>
</rm:provider>
</rm:searchPath>

And then defined corresponding locators. So far so good, but when trying
to resolve a cquery using this rmap, I run into errors like this:

> ERROR [0002] : No suitable provider for component org.eclipse.etrice.generator:osgi.bundle was found in resourceMap file:/D:/git/org.eclipse.etrice.releng.feature/etrice-git.rm ap
> ERROR [0002] : No suitable provider for component org.eclipse.etrice.generator:osgi.bundle was found in searchPath etrice.plugins
> ERROR [0002] : Rejecting provider git({0}/org.eclipse.etrice,plugins/{1}[C:/Users/Daniel/works pace/default/org.eclipse.etrice,plugins/org.eclipse.etrice.g enerator]): No component match was found
> ERROR java.io.IOException: Cannot lock file C:\Users\Daniel\workspace\default\org.eclipse.etrice\.git\co nfig
> E [0002] : Rejecting provider git({0}/org.eclipse.etrice,tests/{1}[C:/Users/Daniel/workspa ce/default/org.eclipse.etrice,tests/org.eclipse.etrice.gener ator.tests]): No component match was found: java.io.IOException: Cannot lock file C:\Users\Daniel\workspace\default\org.eclipse.etrice\.git\co nfig
> ERROR [0002] : No suitable provider for component org.eclipse.etrice.ui.behavior:osgi.bundle was found in resourceMap file:/D:/git/org.eclipse.etrice.releng.feature/etrice-git.rm ap
> ERROR [0002] : No suitable provider for component org.eclipse.etrice.ui.behavior:osgi.bundle was found in searchPath etrice.plugins
> ERROR [0002] : Rejecting provider git({0}/org.eclipse.etrice,plugins/{1}[C:/Users/Daniel/works pace/default/org.eclipse.etrice,plugins/org.eclipse.etrice.u i.behavior]): No component match was found
> ERROR java.io.IOException: Cannot lock file C:\Users\Daniel\workspace\default\org.eclipse.etrice\.git\co nfig

Looks to me like buckminster is trying to lock the local repository
multiple times. The local clone was not present, I intended to let
buckminster take care of this. Does the git reader not support multiple
searchPaths using the same repository?

Here's the feature versions I used:

> Buckminster - Core 1.3.1.r11666 org.eclipse.buckminster.core.feature.feature.group
> Buckminster - Git Support 1.0.360.r11654 org.eclipse.buckminster.git.feature.feature.group
> Buckminster - PDE support 1.2.1.r11660 org.eclipse.buckminster.pde.feature.feature.group
> Eclipse EGit (Incubation) 0.10.1 org.eclipse.egit.feature.group
> Eclipse JGit (Incubation) 0.10.1 org.eclipse.jgit.feature.group

Any help would be appreciated.

Regards,
Daniel
Re: [git-reader] Materializing components from different sub-paths of the same repository [message #654862 is a reply to message #654845] Thu, 17 February 2011 13:22 Go to previous messageGo to next message
Thomas Hallgren is currently offline Thomas Hallgren
Messages: 3229
Registered: July 2009
Senior Member
On 2011-02-17 13:21, Daniel Weber wrote:
>
> Looks to me like buckminster is trying to lock the local repository multiple times. The local clone was not present, I
> intended to let buckminster take care of this. Does the git reader not support multiple searchPaths using the same
> repository?
>
AFAIK, it has never been tested so there's a distinct possibility that it doesn't work just yet. Our git support is
still in it's infancy and it builds on egit/jgit which is still in incubation and moving forward rapidly. It's all for
the good but help finding bugs, attached samples that provoke errors, and of course patches fixing such errors ;-) is
greatly appreciated. Nothing mandatory of course, but the more the better.

Please enter a bugzilla for this and attach as much info as possible.

Thanks,
Thomas Hallgren
Re: [git-reader] Materializing components from different sub-paths of the same repository [message #654897 is a reply to message #654862] Thu, 17 February 2011 14:40 Go to previous messageGo to next message
Carsten Reckord is currently offline Carsten Reckord
Messages: 139
Registered: July 2009
Senior Member
On 17.02.2011 14:22, Thomas Hallgren wrote:
> On 2011-02-17 13:21, Daniel Weber wrote:
>>
>> Looks to me like buckminster is trying to lock the local repository multiple times. The local clone was not present, I
>> intended to let buckminster take care of this. Does the git reader not support multiple searchPaths using the same
>> repository?
>>
> AFAIK, it has never been tested so there's a distinct possibility that it doesn't work just yet. Our git support is
> still in it's infancy and it builds on egit/jgit which is still in incubation and moving forward rapidly. It's all for
> the good but help finding bugs, attached samples that provoke errors, and of course patches fixing such errors ;-) is
> greatly appreciated. Nothing mandatory of course, but the more the better.
>
> Please enter a bugzilla for this and attach as much info as possible.
>
> Thanks,
> Thomas Hallgren

I can confirm this. When initially resolving a query where the git
repository is not yet present locally, concurrency issues ensue.

I've avoided this so far by setting the parallel resolver threads to 1.

IIRC, the problem I identified in the past is that the first resolver thread
starts cloning the repository, then the second thread tries to resolve
something else from it, sees that there already is a repository (and doesn't
know about or gets blocked by the other resolver still cloning) and tries to
checkout from it.
Re: [git-reader] Materializing components from different sub-paths of the same repository [message #655210 is a reply to message #654897] Fri, 18 February 2011 16:24 Go to previous messageGo to next message
Daniel Weber is currently offline Daniel Weber
Messages: 51
Registered: July 2009
Member
On 17.02.2011 15:40, Carsten Reckord wrote:
> I've avoided this so far by setting the parallel resolver threads to 1.

That seems to be a valid workaround, worked for me, too. Thanks!

> IIRC, the problem I identified in the past is that the first resolver thread
> starts cloning the repository, then the second thread tries to resolve
> something else from it, sees that there already is a repository (and doesn't
> know about or gets blocked by the other resolver still cloning) and tries to
> checkout from it.

So has a bug been filed already?

Cheers,
Daniel
Re: [git-reader] Materializing components from different sub-paths of the same repository [message #655400 is a reply to message #655210] Sun, 20 February 2011 15:06 Go to previous messageGo to next message
Carsten Reckord is currently offline Carsten Reckord
Messages: 139
Registered: July 2009
Senior Member
On 18.02.2011 17:24, Daniel Weber wrote:
> On 17.02.2011 15:40, Carsten Reckord wrote:
>> I've avoided this so far by setting the parallel resolver threads to 1.
>
> That seems to be a valid workaround, worked for me, too. Thanks!
>
>> IIRC, the problem I identified in the past is that the first resolver thread
>> starts cloning the repository, then the second thread tries to resolve
>> something else from it, sees that there already is a repository (and doesn't
>> know about or gets blocked by the other resolver still cloning) and tries to
>> checkout from it.
>
> So has a bug been filed already?

No, I meant to file one and look into it further, too, but I haven't come
around to either yet. So feel free to go ahead and file one.
Re: [git-reader] Materializing components from different sub-paths of the same repository [message #657616 is a reply to message #655400] Thu, 03 March 2011 13:40 Go to previous message
Daniel Weber is currently offline Daniel Weber
Messages: 51
Registered: July 2009
Member
See https://bugs.eclipse.org/bugs/show_bug.cgi?id=338810
Previous Topic:Product Build with Buckminster
Next Topic:Problem with rmap file?
Goto Forum:
  


Current Time: Thu Sep 18 16:03:16 GMT 2014

Powered by FUDForum. Page generated in 0.02393 seconds
.:: Contact :: Home ::.

Powered by: FUDforum 3.0.2.
Copyright ©2001-2010 FUDforum Bulletin Board Software