diff options
| author | Christopher Faylor <me@cgf.cx> | 2011-12-04 17:58:24 +0000 |
|---|---|---|
| committer | Christopher Faylor <me@cgf.cx> | 2011-12-04 17:58:24 +0000 |
| commit | 8f937028d9b8d387cf35b64d92056241bffabd78 (patch) | |
| tree | 2ed73616733809fc902b1e33bc2370a0468ba142 /winsup/cygwin/fhandler_socket.cc | |
| parent | 69864e48cb0ef1efd1ac29a80b34bbc9c604533b (diff) | |
| download | cygnal-8f937028d9b8d387cf35b64d92056241bffabd78.tar.gz cygnal-8f937028d9b8d387cf35b64d92056241bffabd78.tar.bz2 cygnal-8f937028d9b8d387cf35b64d92056241bffabd78.zip | |
* exceptions.cc (exception::handle): Drop abbreviation for "exception" since I
never remember what it stands for.
(sig_handle_tty_stop): Remove obsolete call to sig_handle_tty_stop.
(_cygtls::call_signal_handler): Rework to grab signal information from
_main_tls if none is set for _my_tls. Try harder to keep thread locked.
(reset_signal_arrived): Delete.
* fhandler.cc (fhandler_base_overlapped::wait_overlapped): Use new cygWFMO call
to wait for an event + standard cygwin stuff. Modify debug output to
acccomodate new function.
* fhandler_console.cc (fhandler_console::read): Replace WaitForMultipleObjects
with cygWFMO.
* fhandler_socket.cc (get_inet_addr): Add comment.
* gendef (_sigdelayed): Remove call to reset_signal_arrived.
* sigproc.cc (_cygtls::signal_exit): Don't close my_readsig here unless we're
in the signal thread.
(create_signal_arrived): Create signal_arrived as auto-reset so that only one
thread is woken when a signal arrives.
* sigproc.h (cygWFMO): New function.
(reset_signal_arrived): Delete declaration.
Diffstat (limited to 'winsup/cygwin/fhandler_socket.cc')
| -rw-r--r-- | winsup/cygwin/fhandler_socket.cc | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/winsup/cygwin/fhandler_socket.cc b/winsup/cygwin/fhandler_socket.cc index fff1083c3..17d98dc38 100644 --- a/winsup/cygwin/fhandler_socket.cc +++ b/winsup/cygwin/fhandler_socket.cc @@ -127,6 +127,8 @@ get_inet_addr (const struct sockaddr *in, int inlen, some greedy Win32 application. Therefore we should never wait endlessly without checking for signals and thread cancel event. */ pthread_testcancel (); + /* Using IsEventSignalled like this is racy since another thread could + be waiting for signal_arrived. */ if (IsEventSignalled (signal_arrived) && !_my_tls.call_signal_handler ()) { |