You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There was a theoretical deadlock possible when an external thread would have filled the
channel while a task on the event loop tries to schedule, potentially blocking the main
thread.
Additionally, when multiple requests schedule in parallel before MAIN_TID is
initialized, they could deadlock (because they have to go via nginx_notify — although
they come from main thread, we didn't capture MAIN_TID yet). Making the channel
unbounded prevents that. After MAIN_TID is initialized, schedule will always .run(), not
schedule.
0 commit comments