Issue1049

classification
Title: tes_socket_ssl test_rude_shutdown deadlocks
Type: Severity: normal
Components: Core Versions:
process
Status: open Resolution:
Dependencies: Superseder:
Assigned To: amak Nosy List: amak, pjenvey
Priority: Keywords:

Created on 2008-06-08.01:00:31 by pjenvey, last changed 2008-06-11.18:44:01 by amak.

Files
File name Uploaded Description Edit Remove
SSLConnectTest.java amak, 2008-06-11.18:44:00
Messages
msg3242 (view) Author: Philip Jenvey (pjenvey) Date: 2008-06-08.01:00:29
This test is currently disabled because the SSL handshaking appears to 
deadlock when the server has rudely closed the connection
msg3274 (view) Author: Alan Kennedy (amak) Date: 2008-06-11.18:44:00
This appears to be a java issue.

The attached java file implements the same series of java calls as the
jython test_rude_shutdown; the test hangs on the
ssl_socket.startHandshake().

This happens regardless if the socket was created with java.nio (as the
jython socket module does) or with java.net (sometimes the socket
creation method can affect later behaviour).

Given that this bug can be recreated with pure java equivalent to the
failing jython, it is likely that this is a java bug.

If it is a bug at all; I'm a little dubious about an SSL client
handshaking with a non-SSL server socket. But if the test passes on
cpython, then it is obviously a valid thing to do, on some platforms at
least.

I'd like to do a little more reading before raising a bug with Sun about
this. Meantime, all thoughts and opinions welcome.

BTW, I tested this on Ubuntu and Sun JDK 1.5.0_13. Will test on other
platforms also.
History
Date User Action Args
2008-06-11 18:44:01amaksetfiles: + SSLConnectTest.java
messages: + msg3274
2008-06-08 01:00:31pjenveycreate