[
https://issues.apache.org/jira/browse/SLING-2086?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Carsten Ziegeler resolved SLING-2086.
-------------------------------------
Resolution: Fixed
> Potential endless loop in jcr install
> -------------------------------------
>
> Key: SLING-2086
> URL: https://issues.apache.org/jira/browse/SLING-2086
> Project: Sling
> Issue Type: Bug
> Components: Installer
> Affects Versions: JCR Installer 3.0.4
> Reporter: Carsten Ziegeler
> Assignee: Carsten Ziegeler
> Fix For: JCR Installer 3.1.0
>
>
> The log looks like the following:
> ....
> 12.05.2011 13:03:27.747 *INFO* [JcrInstaller.4]
> org.apache.sling.installer.provider.jcr.impl.WatchedFolder Watching folder
> /libs/mcm/install (priority 100)
> 12.05.2011 13:03:27.748 *INFO* [JcrInstaller.4]
> org.apache.sling.installer.provider.jcr.impl.WatchedFolder Watching folder
> /libs/crxde/install (priority 100)
> 12.05.2011 13:03:27.750 *INFO* [JcrInstaller.4]
> org.apache.sling.installer.provider.jcr.impl.WatchedFolder Watching folder
> /apps/geometrixx/install (priority 200)
> 12.05.2011 13:03:27.752 *INFO* [JcrInstaller.4]
> org.apache.sling.installer.provider.jcr.impl.JcrInstaller Removing resource
> from
> OSGi installer (folder deleted): []
> This repeats every 500ms.
> The WatchedFolder.rescanTimer is set, but all WatchedFolders.needsScan()
> return
> false. That leads to the endless loop, since in JcrInstaller.runOneCycle the
> variable scanWf is always true, but for every watched folder needsScan()
> returns
> false. Therefore the rescanTimer is never reset, but updateFoldersList() is
> called every 500ms.
> Probably this is caused by the following timing (for example if you delete
> /apps/geometrixx/install):
> 1. The background thread executes runOnCycle in JcrInstaller up to line
> 446
> (at this time the rescanTimer of the WatchedFolders is not set, but
> updateFoldersListTimer.expired() is true)
> 2. The event listener thread calls onEvent of the WatchedFolder for
> /apps/geometrixx/install. This leads to a cal of scheduleScan() which itself
> sets the rescanTimer and also sets needScan to true.
> 3. Back in the background thread updateFoldersList() is executed (line 453
> of
> JcrInstaller). In this method it detects that one of the watchedFolder have
> been
> removed and therefore indirectly resets the needScan of the WatchedFolder
> /apps/geometrixx/install back to false (line 385). Unfortunately it doesn't
> reset the rescanTimer.
> 4. The next time when runOneCycle is executed, the rescanTimer has exired
> ->
> scanWF is true. Unfortunately every watchedFolder returns false for
> needsScan().
> Now you are in a loop, because the rescanTimer will not be reset.
> The main problem is, that needsScan and rescanTimer.reset/set must always be
> set
> in parallel, but this is not true for the described usecase.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
