[
https://issues.apache.org/jira/browse/UIMA-5058?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Burn Lewis resolved UIMA-5058.
------------------------------
Resolution: Fixed
Restored the original check-for-matching-descriptions approach but added a
cache holding the URLs that were parsed while creating each cached metadata
object. Now when a cached object is used the list of already-processed
descriptors is updated so that creation of an identical pipeline will get the
same set of metadata from the cache.
> Scaled out pipelines can falsely trigger the error: Illegal adding of
> additional MetaData after CASes have been defined
> -----------------------------------------------------------------------------------------------------------------------
>
> Key: UIMA-5058
> URL: https://issues.apache.org/jira/browse/UIMA-5058
> Project: UIMA
> Issue Type: Bug
> Components: Core Java Framework
> Affects Versions: 2.8.1SDK
> Reporter: Burn Lewis
> Assignee: Burn Lewis
> Fix For: 2.9.0SDK
>
>
> When UIMA-AS scales out a synchronous UIMA pipeline it sequentially creates
> separate AEs in separate threads, sharing the resource manager. Creation of
> the 2nd (identical) AE will fail with "Illegal adding of additional MetaData
> after CASes have been defined" if a descriptor imports 2 type-system files,
> one of which is included in the other. The check fails to recognize that the
> types loaded by the 2nd AE are already in the resource manager because of the
> way type-systems are cached.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
