Shuo Chen
Shuo Chen
muduo 不会出现你说的情况，`wakeup()` 会正确地及时唤醒 `epoll_wait()`。 假设主线程和 `EventLoop::loop()` 位于两个线程，如果主线程调用了 `EventLoop::runInLoop()`，它会调用 `EventLoop::wakeup()`，往 `wakeupFd_` 写一次数据。那么当 EventLoop 线程进入到 `epoll_wait(2)` 时，`wakeupFd_` 是可读的，`epoll_wait(2)` 会立刻返回，并不会阻塞 `kPollTimeMs` 这么长的时间。
欢迎提 Pull Requests.
Please fix travis-ci first, put `apt-get install` in `.travis.yml`.
Did you upgrade or downgrade protobuf library / compiler recently? Or installed protobuf from source instead of 'apt get'? Ubuntu 18.04 comes with protobuf 3.0.0: https://packages.ubuntu.com/bionic/protobuf-compiler Try ```sh $ dpkg...
Unlikely. I suggest you set up a HTTPS server as reverse proxy, and implement business logic as a FastCGI backend using muduo.
Please provide a [SSCCE](http://sscce.org/): Short, Self Contained, Correct (Compilable), Example.
不一定出现不要紧。至少要有一个完整的能运行的程序，不然光凭文字描述我很难复现你遇到的场景。 另外，我不认识 `yuto::net` 这个 namespace，如果你用的不是 muduo 库本尊，请考虑把 issue 提到 `yuto` 库那里。
Why do you think this PR improves performance? And data? Quite the opposite, I think this PR harms performance because it shrinks too much.
The original code guarantees that the buffer is at least 1024B, your new code doesn't. That mean your code could need more memory allocations when buffer grows large.
Can you show performance testing data that justifies this PR?