[jira] [Commented] (VFS-834) input streams prematurely closed

["Ivan Bella (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/VFS-834?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17704330#comment-17704330 ] 

Ivan Bella commented on VFS-834:
--------------------------------

I will be putting up a pull request momentarily for this ticket.

> input streams prematurely closed
> --------------------------------
>
>                 Key: VFS-834
>                 URL: https://issues.apache.org/jira/browse/VFS-834
>             Project: Commons VFS
>          Issue Type: Bug
>    Affects Versions: 2.4
>            Reporter: Ivan Bella
>            Priority: Major
>             Fix For: 2.10.0
>
>
> This was discovered while using Spring with an underlying VFSClassLoader.  If one created a Spring context uses "classpath*:/MyFile.xml" then everything works smoothly.  However if using "classpath:/MyFile.xml" then it would result in a failure to get the contents of the file manifested as a sax parse exception at line 0, character 0.  The workaround was easy (use classpath*:/...) however trying to figure out why this was occuring took some time.
> The underlying reason was that the VFSClassLoader.loadResource method changed in version 2.4 to wrap the FileObject returned by resolveFile in a try in this commit:
> {noformat}
> commit 3b2de84556b88a51af42d38b99ce6bf81b970cb5
> Author: Gary Gregory <ga...@gmail.com>
> Date:   Thu Jun 27 11:23:14 2019 -0400    [VFS-719] Add methods to get the contents of file objects as strings.{noformat}
> As it turns out closing a FileObject subsequently closes input streams created within the same thread for that same file.
> The reason this broke Spring is that in one of those paths, an object was created that did a lazy loading of the resource which ended up closing an input stream that was still in use by that same thread (yes spring opens and reads the file twice, once for validation and once to resolve the configured beans).
> It does not appear there is a good reason to close the FileObject in this case.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)