[jira] [Reopened] (NET-73) [net][PATCH] TelnetInputStream.read hangs when socket data ends in a command sequence

["Sebb (Reopened) (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/NET-73?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sebb reopened NET-73:
---------------------


The 2.x (and 3.x) line of code was patched in r633798, and I've checked that the same patch is also in the current 3.0.1 code.

==

I think the stack trace shows that the code is waiting to close a previous input stream as part of a connect.

This suggests that the problem may lie in the way the Telnet classes are being used.
Can you provide sample code that demonstrates the problem?
                
> [net][PATCH] TelnetInputStream.read hangs when socket data ends in a command sequence
> -------------------------------------------------------------------------------------
>
>                 Key: NET-73
>                 URL: https://issues.apache.org/jira/browse/NET-73
>             Project: Commons Net
>          Issue Type: Bug
>         Environment: Operating System: other
> Platform: All
>            Reporter: Rob Hasselbaum
>         Attachments: cmd-seq-hang.patch
>
>
> Background: If one calls TelnetInputStream.read() in single-threaded mode (no 
> reader thread) and there is no data immediately available, the call blocks on 
> a socket read. When data starts to arrive, the stream adds all the available 
> bytes to its internal queue before returning the first one to the caller. To 
> do this, it calls __read() in a loop for as long as there are bytes available. 
> The __read() method returns the first byte of "user data" from the socket. If 
> __read() encounters a Telnet command sequence (IAC, WILL, WONT, DO, DONT, 
> etc.), it handles the negotiation transparently and then returns the first 
> byte of user data.
> In most cases, this works fine, but a problem arises if a chunk of data from 
> the remote host ends in a Telnet command sequence. When that happens, the 
> TelnetInputStream.read() method hangs, even though it may have already 
> acquired some user data. This is because it calls __read() in a loop as long 
> as super.available() returns true. But if the remaining data from the socket 
> consists entirely of Telnet commands, __read() will process those AND THEN 
> BLOCK waiting for user data.
> Just checking super.available() is not sufficient. We should continue the loop 
> only if there are bytes of USER DATA still available from the socket. Not 
> doing this can cause the client to wait indefinitely.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira