[Syncevolution-issues] [Bug 7555] syncevo-dbus-server: CTRL-C/SIGINT/SIGTERM

(In reply to comment #7) > (In reply to comment #6) > > While testing with test-dbus.py I noticed that the server was not shut down > > reliably after running a sync. I think that was a timing issue: if the SIGTERM > > arrived after running a sync and before returning into the syncevo-dbus-server > > main loop, g_main_loop_quit() would be called outside of the loop, thus doing > > nothing, so that the signal was ignored. > If loop is running, quit_loop won't do anything about loop. I add below code > to check while in the DBusServer::run(). You can find them in the branch > 'origin/yongsheng'. > > > --- a/src/syncevo-dbus-server.cpp > +++ b/src/syncevo-dbus-server.cpp > @@ -2561,12 +2561,6 @@ void DBusServer::run() > // running a sync and quitting; no active session, so quit > break; > } > + /** check whether receiving CTRL-C/SIGINT/SIGTERM and aborting > syncevo-dbus-server */ > + SuspendFlags flags = SyncContext::getSuspendFlags(); > + if(flags.state == SuspendFlags::CLIENT_SUSPEND || > + flags.state == SuspendFlags::CLIENT_ABORT ) { > + break; > + } What about the case that we are in DBusServer::run() and outside of SyncContext::doSync()? In that case the active signal handler is the niam() function in syncevo-dbus-server. It doesn't set SyncContext::getSuspendFlags(). There's another problem related to this: Session::abort/suspend() do not tell the DBusTransportAgent to return control. Because of that, the user cannot abort a session when no messages are coming in. The timeout mechanism would help, but isn't implemented yet, so abort() really has no effect at all. Are you done with the code? I'm working on these other aspects of the code at the moment, so I should take over and try fix these problems together.

(In reply to comment #7) > (In reply to comment #6) > > While testing with test-dbus.py I noticed that the server was not shut down > > reliably after running a sync. I think that was a timing issue: if the SIGTERM > > arrived after running a sync and before returning into the syncevo-dbus-server > > main loop, g_main_loop_quit() would be called outside of the loop, thus doing > > nothing, so that the signal was ignored. > If loop is running, quit_loop won't do anything about loop. I add below code > to check while in the DBusServer::run(). You can find them in the branch > 'origin/yongsheng'. > > > --- a/src/syncevo-dbus-server.cpp > +++ b/src/syncevo-dbus-server.cpp > @@ -2561,12 +2561,6 @@ void DBusServer::run() > // running a sync and quitting; no active session, so quit > break; > } > + /** check whether receiving CTRL-C/SIGINT/SIGTERM and aborting > syncevo-dbus-server */ > + SuspendFlags flags = SyncContext::getSuspendFlags(); > + if(flags.state == SuspendFlags::CLIENT_SUSPEND || > + flags.state == SuspendFlags::CLIENT_ABORT ) { > + break; > + } What about the case that we are in DBusServer::run() and outside of SyncContext::doSync()? In that case the active signal handler is the niam() function in syncevo-dbus-server. It doesn't set SyncContext::getSuspendFlags().

There's another problem related to this: Session::abort/suspend() do not tell the DBusTransportAgent to return control. Because of that, the user cannot abort a session when no messages are coming in. The timeout mechanism would help, but isn't implemented yet, so abort() really has no effect at all. Are you done with the code? I'm working on these other aspects of the code at the moment, so I should take over and try fix these problems together.

abort() also doesn't shut down the connection, therefore the Session gets stuck in its shutdown waiting for the peer. Fixing it... We really need tests for these API calls ;-}