[
https://issues.apache.org/jira/browse/NET-220?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sebb updated NET-220:
---------------------
Attachment: TFTPServerPathTest.patch
TFTPServer.patch
Patch to TFTPServer changes all instance variables to final or volatile.
There did not seem to be a use-case for changing the logging destinations after
construction, so rather than making these volatile (or adding synchronisation)
I removed the set() methods.
The method used to start the server is called start(), but could of course be
renamed if required.
> TFTPServer is not threadsafe
> ----------------------------
>
> Key: NET-220
> URL: https://issues.apache.org/jira/browse/NET-220
> Project: Commons Net
> Issue Type: Bug
> Affects Versions: 2.0
> Reporter: Sebb
> Attachments: TFTPServer.patch, TFTPServerPathTest.patch
>
>
> The TFTPServer class is not threadsafe.
> There are several reasons for this:
> Ctors call private launch() method which creates a thread and start()s it.
> This publishes the instance before it has been fully constructed.
> Various instance fields are not thread-safe - e.g. serverReadDirectory_ is
> neither final nor volatile, and accesses to it are not synchronized.
> Although it is only written by the ctors, the lack of synch means that the
> field may not be visible to other threads.
> As far as possible, instance fields should be made final - this guarantees
> that the field will be visible to other threads.
> One of the shutdown_ variables is volatile (so is threadsafe) but the
> shutdown_ variable in the nested TFTPTransfer class is not, yet is accessed
> from multiple threads.
> The access to the variable serverException is not synch; should probably be
> volatile.
> Patch to follow.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.