On Nov 18, 2004, at 7:17 PM, James Y Knight wrote:
On Nov 18, 2004, at 3:58 AM, Bob Ippolito wrote:
Very test-failness. I haven't seen a situation where cfreactor doesn't actually work in practice since it was changed to handle the GIL correctly.
There are two serious issues I know about: 1) See "survive a re-entrant call to .resumeProducing, which happens under the cf reactor". That should definitely not happen. Unfortunately, there is no test case testing this at the moment.
If someone files a bug with a stack trace, I'm sure this would be easy to diagnose and fix. I have seen no such thing, so I can't fix it. It's probably resumeProducing calling doWrite calling resumeProducing, or something.. but I'm not going to attempt a fix-by-guess.
2) Half-close support is unimplemented.
There's three solutions to that: - Someone (other than me) can implement it. - Someone (other than me) can refactor the default file descriptor implementations so that cfreactor can be implemented cleaner and correctly share all this half-closing code. If this happens, I'll refactor cfreactor accordingly. - The half-close interface can be made optional. -bob