| Summary: | blocking with oneway | ||
|---|---|---|---|
| Product: | TAO | Reporter: | Nanbor Wang <bala> |
| Component: | ORB | Assignee: | Simon McQueen <sm> |
| Status: | RESOLVED DUPLICATE | ||
| Severity: | normal | CC: | gruetzmacher |
| Priority: | P3 | ||
| Version: | 1.2.1 | ||
| Hardware: | x86 | ||
| OS: | Windows NT | ||
| Attachments: | Sample | ||
Created attachment 107 [details]
Sample
Just accepting the bug Taken. Accepting. This problem no longer exists following the fix to bug #1222. The deceptively named 'server' in the supplied example now receives an appropriate system exception when the 'client' (receiving the oneway calls) is killed. Thanks to Carlos O'Ryan for the tip off re this. One point of minor note is that using the method described in the bug report to close the connection (blocking the 'client' with a breakpoint and then ending the debug session) causes the process to exit with a horrorific exception in VC++ the *second* time you do it. This seems to be a peculiarity of the IDE debugger only as it does not occur when simulating the same situation any other way. *** This bug has been marked as a duplicate of 1222 *** |
TAO VERSION: 1.2.1 ACE VERSION: 5.2.1 HOST MACHINE and OPERATING SYSTEM: WinNT 4.0 SP 6 COMPILER NAME AND VERSION (AND PATCHLEVEL): MSVC 6.0 SP 5 AREA/CLASS/EXAMPLE AFFECTED: My application is hanging in a oneway call. DESCRIPTION: I've written a simple server/client set (already send to this list some times). Server is waiting for a client which is registering it's callback function (oneway) which will be called endless. REPEAT BY: Unpack attached example. Open server and client project in separate Visual Studio instances. Compile (client with debug infos). Set WorkingDir of both applications to the same Dir (in the example the dir with the svc.conf . Start Server (the IOR string will be written to a file in the working dir) - Wait for message "Ready on ORB". Start Client (it should read the ior file and try to connect the server). Now the server should write "Send Message XXX" and the client too - the communication is established successfull. Set a breakpoint somewhere in the client callback implemetation CallTest (client.cpp). The server keeps sending messages but only further around 1900 messages - than the server stops. If you now remove the breakpoint and continue the execution the server go back to sending. But if I stop the debugging of the client instead of continue execution the server is unrecoverable blocked. I can restart the client and I see "Client connected" but the server sending thread is hanging in ace\os.i:6031 : ACE_SOCKCALL_RETURN (::select (width, (ACE_FD_SET_TYPE *) rfds, (ACE_FD_SET_TYPE *) wfds, (ACE_FD_SET_TYPE *) efds, timep), int, -1); SAMPLE FIX/WORKAROUND: see attached.