labath wrote: > BTW, I have figured out a side effect. MainLoopWindows uses events created by > WSACreateEvent(). WSACreateEvent() creates the `manual-reset` event. But > WSAResetEvent() is used only for m_trigger_event. All other events (READ, > ACCEPT, CLOSE) are never reset. I have no idea how it worked before. But I > got thousands of error messages using this patch with lldb-server after first > connection: > > ``` > AcceptSocket(): A non-blocking socket operation could not be completed > immediately > ``` > > I have added `WSAResetEvent(KV.second.event);` before > `ProcessReadObject(KV.first);` in MainLoopWindows.cpp to fix it. It seems > this issue must be fixed before or with this patch, otherwise it will break > the Windows build.
Oops. Thanks for catching that. I think have a pretty good idea of how this worked before. For accepts, we were always creating a fresh MainLoop instance (and a WSAEvent to go along with it), so this wasn't an issue. And for reads, we probably were just spinning in a busy loop after the first read event. I've created #107061 to fix that. https://github.com/llvm/llvm-project/pull/106955 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
