Thanks to Tim Becker for pointing this out.

Reported here:
http://groups.google.com/group/nodejs-dev/browse_thread/thread/9e063d0938f99879

Would be good to test this somehow...

Reported by Kadir Pekel <kadirpekel@gmail.com>

Made this test in response to this thread:
http://groups.google.com/group/nodejs/browse_thread/thread/f82835007a277de2/
But Node appears to be working correctly.

New() methods should be invoked as constructors, not regular functions.
Corner cases like Script::New() may cause a SIGSEGV when the GC is run.

More details: http://groups.google.com/group/nodejs/browse_thread/thread/a7e5db68d4cd6356

onend and ondata was cleaning on parser end

Original patch c/o Evan Larkin <evan.larkin.iit@gmail.com>

This reverts commit e2274412.

Build fails on mac

rnewson found a good bug in keep-alive. we were only using the request
headers we send to enable/disable keep-alive but when the server sends
Connection: close we need to close down the connection regardless.

I wrote up a patch the Robert verified makes all his test client code work
now and I also added a new unittest for it.

Leads to lots of builds with out-of-memory.

Updated extensions tests

This patch removes require.async from nodejs.

1. It complicated the code unnecessarily.
2. Everyone uses sync require anyway.
3. It's got a lot of weird edge cases when mixed with sync require.
4. It is many months behind the commonjs spec anyhow.