The module loading code is looking extremely messy. It needs a refactor.

I don't want users to have to think about coroutine safety.

http://thread.gmane.org/gmane.comp.lang.javascript.nodejs/2468/focus=2603

This is to reduce our dependency on wait(). For some reason this patch
affects the timer test:

  % ./node test/mjsunit/test-timers.js
  diff: 989
  diff: 989
  diff: 1989
  diff: 2989

Previously it showed:

  % ./node test/mjsunit/test-timers.js
  diff: 1000
  diff: 1000
  diff: 2000
  diff: 3000

I'm not sure what caused this change, and it's rather disturbing. However I
want to remove wait() as soon as possible and so am pushing this patch
through.

The module loading code is becoming increasingly ugly - this patch has not
helped. A refactor needs to be done soon.

I hope I won't spend all my time now maintaining this file...

Why doesn't WAF just search LD_LIBRARY_PATH?

- Rename fs.cat to fs.readFile
- Move file.write to fs.writeFile
- Allow strings for the flag argument to fs.open
  ("r", "r+", "w", "w+", "a", "a+")
- Remove the unused 'File' module / class

- sendBody() renamed to write()
- 'body' event renamed to 'data'
- 'complete' event renamed to 'end'

And don't use --debugger_auto_break v8 flag, it seems unnecessary.

Bug reported by James Golick.

Unnecessary

This reverts commit 896cef23.

This fixes crash that used to happen when Eclipse debugging session hits the
breakpoint that is set either in timer or connection callback.

Failure introduced in 80ace36e