[
https://issues.apache.org/jira/browse/JCR-3958?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15230047#comment-15230047
]
Takahito Kikuchi commented on JCR-3958:
---------------------------------------
[[email protected]], I checked the code of your fix, then there does not seem
to be any problems.
And I verified, then it works fine.
It was not easy to do async upload in order to return {{null}} of {{remove}}
purposely, so instead I set a specified sentence such as 'abc' onto {{remove}
of {{Localcache}} you implemented this time so that we can check just the
behavior.
{code:java}
//if (asyncUploadCache.hasEntry(fileName, false)) {
if (asyncUploadCache.hasEntry(fileName, false) ||
fileName.contains("abc")) {
{code}
To compare your fix with the previous, I built the previous version with this
same condition of specified sentence. The result is I could reproduce endless
loop happened.
Next, I tried with your fix then it works fine. {{LocalCache}} with your fix
can detect the entry with the specified sentence once only, then skip it to
next. Also, I confirmed the cache purge runs well each time it reaches
cachePurgeTrigFactor. That looks good.
> Endless loop until async upload completes
> -----------------------------------------
>
> Key: JCR-3958
> URL: https://issues.apache.org/jira/browse/JCR-3958
> Project: Jackrabbit Content Repository
> Issue Type: Bug
> Affects Versions: 2.12.0
> Reporter: Takahito Kikuchi
> Assignee: Unico Hommes
>
> This is {{LocalCache#PurgeJob()}} issue.
> If the entry object in cache is removed successfully by cache.remove(),
> cache.entrySet would be decreased then next entry will be retrieved and
> removed. However, if the entry object is not removed due to the failure of
> removing the file during async upload, the entry in cache would not be
> removed, then the next entry would be the same entry again by the iterator
> because the iterator is not changed still. Consequently, that causes endless
> loop until async upload of the file completes. If the file is larger, the
> endless loop will continue for a longer time. It's harmful.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
