[
https://issues.apache.org/jira/browse/FLINK-2448?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14704229#comment-14704229
]
ASF GitHub Bot commented on FLINK-2448:
---------------------------------------
Github user sachingoel0101 commented on the pull request:
https://github.com/apache/flink/pull/1031#issuecomment-132871675
Was there a specific reason to not generate fresh environments in the first
place? Or was it just to minimize object creations?
Here's what I propose though: Since there is already a `clearSinks`
argument to `createProgramPlan`, we can use that as an indicator that the cache
file list has to be cleared too, since it indicates a new Execution.
Further, instead of throwing an exception in case a `name` is
re-registered, we check if the `filePath` is actually the same. If not, then we
throw an exception, otherwise just silently ignore it.
> registerCacheFile fails with MultipleProgramsTestbase
> -----------------------------------------------------
>
> Key: FLINK-2448
> URL: https://issues.apache.org/jira/browse/FLINK-2448
> Project: Flink
> Issue Type: Bug
> Components: Tests
> Reporter: Chesnay Schepler
> Priority: Minor
>
> When trying to register a file using a constant name an expection is thrown
> saying the file was already cached.
> This is probably because the same environment is reused, and the cacheFile
> entries are not cleared between runs.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
