Python 3.0 support

[kanaka]

Hopefully 2.X and 3 can be supported in the same code-base or at least by a mostly common module.

[kanaka]

Info about using the same code-base for 2.X and 3.X: http://nedbatchelder.com/blog/200910/running_the_same_code_on_python_2x_and_3x.html

[kanaka]

Here is a patch that adds python 3 and multiprocessing support: https://gist.github.com/976972

The patch drops python2.4 support for now although the same code base is able to support python 2.6 on up. However, the code is able to trigger a futex hang in python 3.1 so I'm not pushing it into the main tree until that is resolved. The futex hang can be triggered by reloading the noVNC vnc.html page repeatedly until it stops responding. This hang is not related to the WebSockets code since this is just loading the page without doing any WebSocket connections.

stracing the hung python process shows the following:
$ sudo strace -p 2402
[sudo] password for foo:
Process 2402 attached - interrupt to quit
futex(0x1469570, FUTEX_WAIT_PRIVATE, 0, NULL^C <unfinished ...>

When hung like that the process cannot be interrupted with Ctrl-C and must be killed directly.

[kanaka]

Include @ysangkok

[kanaka]

Okay, I figured out a workaround to the python 3 futex hang. It appears to be related to the fact that websocket.py was doing it's own SIGCHLD child process reaping. The multiprocessing module does it's own SIGCHLD handling so we only have to do SIGCHLD handling when using os.fork() (under python 2.4).

Support has been committed with c858711.

[ysangkok]

It doesn't work right now because you're importing the "resource" module, which is Unix-only. I believe this is not strictly required though, so it might be easily fixed?

[kanaka]

@ysangkok, this bug is closed (it works with python 3 on Linux). I made a change to make 'resource' optional and noted it in the win32 bug: #2