You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@ant.apache.org by "Daniel Dekany (JIRA)" <ji...@apache.org> on 2009/03/13 00:02:50 UTC

[jira] Created: (IVY-1049) Describe how to setup the caches, especailly for multiple ivysettings.xml-s

Describe how to setup the caches, especailly for multiple ivysettings.xml-s
---------------------------------------------------------------------------

                 Key: IVY-1049
                 URL: https://issues.apache.org/jira/browse/IVY-1049
             Project: Ivy
          Issue Type: Improvement
          Components: Documentation
    Affects Versions: 2.0
            Reporter: Daniel Dekany


When using more the one ivysettings.xml-s inside the same user account (same computer), the result is often a lot of "unknown resolver" errors, and other such mess. The cause is that caches were not set up properly: I guess usually there should be one repository cache per ivysettings.xml, then one dedicated resolution cache per ivysettings.xml too (at least, because if parallel builds use a common ivysettings.xml, then each of them needs its own). These I mostly deduced from mailing list discussions and experimenting, not from the documentation. So, these things should be well described, and then linked from every place that's about wiring ivysettings.xml-s. (Even maybe the "unknown resolver" error messages could contain a hint about what the problem could be.) Also, the users should be warned about the importance of lock strategies. (I know that in principle these could be find out from http://ant.apache.org/ivy/history/latest-release/concept.html#cache, but that description is at best sparse, and the implications are not described. Also it's not on a very well visible place.)

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (IVY-1049) Describe how to setup the caches, especailly for multiple ivysettings.xml-s

Posted by "Jeffrey Sinclair (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/IVY-1049?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12774814#action_12774814 ] 

Jeffrey Sinclair commented on IVY-1049:
---------------------------------------

Eric, to comment on your question: 'What does it matter which resolver actually was used to populate the cache?'... 

Having the cache segregated by resolver is important when you have multiple repositories with the same module naming convention. In the environment I work in we have two repositories: one for production and one for development. We keep the organisation/module/revision the same across the repositories to make it very easy to switch between them: i.e. you simply use a different resolver. Perhaps branches could be used, however we liked the idea that you can look at an ivysettings.xml file or ivy report and clearly see what development dependencies you have, reducing the chance that someone might try to go into production with them.

Similarly a shared cache which is used to resolve 'shadow' dependencies in development needs cache segregation by resolver in order to avoid different shadows trampling over the other's meta data. If Ivy did not store which resolver was used to populate different artifacts in the cache, it would not be able to handle this use-case.

With the exception of aggregate resolvers (such as the chain resolver), I would recommend setting a cache per resolver to avoid any issues with the above scenarios.


> Describe how to setup the caches, especailly for multiple ivysettings.xml-s
> ---------------------------------------------------------------------------
>
>                 Key: IVY-1049
>                 URL: https://issues.apache.org/jira/browse/IVY-1049
>             Project: Ivy
>          Issue Type: Improvement
>          Components: Documentation
>    Affects Versions: 2.0
>            Reporter: Daniel Dekany
>
> When using more the one ivysettings.xml-s inside the same user account (same computer), the result is often a lot of "unknown resolver" errors, and other such mess. The cause is that caches were not set up properly: I guess usually there should be one repository cache per ivysettings.xml, then one dedicated resolution cache per ivysettings.xml too (at least, because if parallel builds use a common ivysettings.xml, then each of them needs its own). These I mostly deduced from mailing list discussions and experimenting, not from the documentation. So, these things should be well described, and then linked from every place that's about wiring ivysettings.xml-s. (Even maybe the "unknown resolver" error messages could contain a hint about what the problem could be.) Also, the users should be warned about the importance of lock strategies. (I know that in principle these could be find out from http://ant.apache.org/ivy/history/latest-release/concept.html#cache, but that description is at best sparse, and the implications are not described. Also it's not on a very well visible place.)

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (IVY-1049) Describe how to setup the caches, especailly for multiple ivysettings.xml-s

Posted by "Eric Sirianni (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/IVY-1049?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12774091#action_12774091 ] 

Eric Sirianni commented on IVY-1049:
------------------------------------

I agree.  I came across this thread while googling for the "unknown resolver" error:

http://www.mail-archive.com/ivy-user@ant.apache.org/msg02262.html

Without knowing the architecture of ivy, I don't understand why the resolver names are stored in with the repository cache.  From a high-level view, it seems that the repository cache is just a cache of JAR files and dependency definitions.  What does it matter which resolver actually was used to populate that cache?

This creates an unwanted coupling between ivysettings.xml files for independent projects.  They must use the same resolver names or face these constant "unknwon resolver" errors.  I think this is more than a documentation issue and should be fixed in ivy if possible.

> Describe how to setup the caches, especailly for multiple ivysettings.xml-s
> ---------------------------------------------------------------------------
>
>                 Key: IVY-1049
>                 URL: https://issues.apache.org/jira/browse/IVY-1049
>             Project: Ivy
>          Issue Type: Improvement
>          Components: Documentation
>    Affects Versions: 2.0
>            Reporter: Daniel Dekany
>
> When using more the one ivysettings.xml-s inside the same user account (same computer), the result is often a lot of "unknown resolver" errors, and other such mess. The cause is that caches were not set up properly: I guess usually there should be one repository cache per ivysettings.xml, then one dedicated resolution cache per ivysettings.xml too (at least, because if parallel builds use a common ivysettings.xml, then each of them needs its own). These I mostly deduced from mailing list discussions and experimenting, not from the documentation. So, these things should be well described, and then linked from every place that's about wiring ivysettings.xml-s. (Even maybe the "unknown resolver" error messages could contain a hint about what the problem could be.) Also, the users should be warned about the importance of lock strategies. (I know that in principle these could be find out from http://ant.apache.org/ivy/history/latest-release/concept.html#cache, but that description is at best sparse, and the implications are not described. Also it's not on a very well visible place.)

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.