Hi Alec, On Thu, May 30, 2019 at 07:59:22PM +0800, Alec Liu wrote: > Hi Willy, > > Here is the updated patch, please help to look into it, thank you!
Thanks. So I've had a look and am seeing something very strange : the connection establishes to the server, the SOCK4 handshake is exchanged, then a connect() is made again from the same connection, I guess that it's just in order to validate that the connection properly established, and from there it works. A few times I saw epoll_wait() wake up in loops with an uncaught write event. I'm currently having a look into this as I suspect it's the same problem and that there is a place where it's not clear that we've completed the handshake. But overall it looks good. I've discussed with Olivier regarding the possible conflicts, and he said that we can merge your patch before his, and that it will not be too much work to adapt so I'll do this. I think that once I've figured what causes the strange things I've seen I'll simply adjust it and merge it. Thanks! Willy
