You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@felix.apache.org by "Richard S. Hall (JIRA)" <ji...@apache.org> on 2010/07/08 22:38:53 UTC

[jira] Issue Comment Edited: (FELIX-2466) Failure to resolve gemini-web bundles

    [ https://issues.apache.org/jira/browse/FELIX-2466?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12886454#action_12886454 ] 

Richard S. Hall edited comment on FELIX-2466 at 7/8/10 4:38 PM:
----------------------------------------------------------------

Technically, it does look like there is a bug in the resolver, although it is not exactly what you might expect. The com.springsource.org.apache.jasper.springsource has this metadata:

    Fragment-Host: com.springsource.org.apache.catalina.springsource;versi
      on="[6.0.20.S2-r5956, 6.0.20.S2-r5956]"

In reality, this declaration is somewhat faulty since it should be using "bundle-version" as the matching attribute for the host, not "version". The Felix resolver obeys the attribute and doesn't find any matching host; however, this appears to not comply with the spec, since the spec only defines one known attribute (i.e., bundle-version) and doesn't specifically allow arbitrary attributes on the Fragment-Host header.

In short, we need to modify our manifest parser to strip unknown attributes from the Fragment-Host header, then it should start to work. Additionally, someone should tell the SpringSource people to use bundle-version instead of version. :-)

      was (Author: rickhall):
    Technically, it does look like there is a bug in the resolver, although it is not exactly what you might expect. The com.springsource.org.apache.jasper.springsource has this metadata:

    Fragment-Host: com.springsource.org.apache.catalina.springsource;versi
      on="[6.0.20.S2-r5956, 6.0.20.S2-r5956]"

In reality, this declaration is somewhat faulty since it should be using "bundle-version" as the matching attribute for the host, not "version". The Felix resolver obeys the attribute and doesn't find any matching host; however, this is appears to not comply with the spec, since the spec only defines one known attribute (i.e., bundle-version) and doesn't specifically allow arbitrary attributes on the Fragment-Host header.

In short, we need to modify our manifest parser to strip unknown attributes from the Fragment-Host header, then it should start to work. Additionally, someone should tell the SpringSource people to use bundle-version instead of version. :-)
  
> Failure to resolve gemini-web bundles
> -------------------------------------
>
>                 Key: FELIX-2466
>                 URL: https://issues.apache.org/jira/browse/FELIX-2466
>             Project: Felix
>          Issue Type: Bug
>          Components: Framework
>    Affects Versions: framework-3.0.1
>         Environment: Gemini Web (1.1.0.M01)
>            Reporter: David Savage
>             Fix For: framework-3.2.0
>
>
> The gemini web bundles available from:
> http://www.eclipse.org/gemini/web/download/
> Fail to resolve in felix:
> 19  INSTALLED  com.springsource.org.apache.catalina.ha.springsource:6.0.20.S2-r5956
> 20  INSTALLED  com.springsource.org.apache.catalina.springsource:6.0.20.S2-r5956
> 21  INSTALLED  com.springsource.org.apache.catalina.tribes.springsource:6.0.20.S2-r5956
> 23  INSTALLED  com.springsource.org.apache.coyote.springsource:6.0.20.S2-r5956
> 26  INSTALLED  com.springsource.org.apache.jasper.springsource:6.0.20.S2-r5956
> 32  INSTALLED  org.eclipse.gemini.web.core:1.1.0.M01
> 33  INSTALLED  org.eclipse.gemini.web.extender:1.1.0.M01
> 34  INSTALLED  org.eclipse.gemini.web.tomcat:1.1.0.M01
> g! start 19
> org.osgi.framework.BundleException: Unresolved constraint in bundle com.springsource.org.apache.catalina.ha.springsource [19]: Unable to resolve 19.0: missing requirement [19.0] package; (&(package=org.apache.catalina)(version>=6.0.20.S2-r5956)(version<=6.0.20.S2-r5956)) - [19.0] package; (&(package=org.apache.catalina)(version>=6.0.20.S2-r5956)(version<=6.0.20.S2-r5956))
> The same set of bundles resolve ok in equinox.
> Guessing it might have something to do with fragments as they're definitely in use.

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