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
Remove event loop error tracking used to trigger a epoll_create1() at a certain error threshold. This tracking was first introduced in [v1.1.0][], triggered by spurious EPOLLERR on I/O watchers
Unconditionally stop I/O watchers that return EPOLLERR or EPOLLHUP, it is up to the watcher callback to clear the error and/or read() the last few bytes from the descriptor. HUP usally means EOF, or that the remote end of a stream or pipe closed, this may also be signaled by read() returning zero
Properly stop and de-register signal and cron/timer watchers from the epoll socket in case of errors, problem introduced in v2.0.0
Mark watcher file descriptor as unintialized on internal error
Fix double-close of cron/timer watchers. Problem triggered when the timer expires and calls uev_exit(), which stops all watchers. When the timer callback returns another call to stop the watcher triggered the double close()
Fix unit test's error handling in watcher callbacks, for reference
Fix use-before-set in cronrun unit test
Make sure to restart unit test's I/O watchers on UEV_ERROR
Make sure to restart example I/O watchers on UEV_ERROR
Properly check for UEV_HUP in unit tests and examples