Changli Gao
Changli Gao
I do think there aren't many users now, so nobody cares.
Yes. It is a issue. But I am afraid that your way doesn't work. There maybe more threads waiting on a fd, and there isn't any reference to the monitoring...
FYI: I just cooked a patch for this issue, and will submit a PR after testing. @winlinvip Is there any unit testing case?
See #5 @lihuiba @winlinvip
@lihuiba reported this issue. I think he will test the PR with his workload. I tested the server in the example directory, and ab reported slight improvements on both QPS...
@lihuiba what is the result？
@lihuiba Are many coroutines monitoring a FD? Is a coroutine monitoring lots of FDs?
@lihuiba Would you provide the testing codes?
It is strange. I ran your code, and got an opposite conclusion. ./server -t 1 -p 1 -b 127.0.0.1:8888 -l . -i Could you run your code again?