I'm still investigating exactly what causes this, but here's a simple test case to show what I'm talking about:
import std.stdio;
import std.socket;
import core.thread;
void main() {
testSocket();
auto t = new Thread(&testSocket);
t.start();
}
shared ushort port = 5000;
public void testSocket() {
try {
auto socket = new TcpSocket();
socket.bind(new InternetAddress("0.0.0.0", port++));
}
catch(SocketException e) {
writefln("Error: %d", e.errorCode);
return;
}
writefln("success!");
}
The output:
success!
Error: 10093
According to http://msdn.microsoft.com/en-us/library/ms740668(VS.85).aspx the 10093 code is WSANOTINITIALISED, which means the following:
Either the application has not called WSAStartup or WSAStartup failed. The application may be accessing a socket that the current active task does not own (that is, trying to share a socket between tasks), or WSACleanup has been called too many times.
WSAStartup was definitely called because the first socket was created successfully. I'm still figuring out how the second one fails.
Comment #1 by debio264 — 2010-11-06T18:35:01Z
Okay, after adding some debugging writeflns to std.socket, the output is now this:
initializing WSA
Initializing socket
success!
cleaning up WSA
Initializing socket
Error: 10093
cleaning up WSA
So the problem is that the module destructor for std.socket is called when the first thread terminates, even though the second thread still needs it.
This is because the module constructor for std.socket is "shared static this()" while the destructor is "static this()". Adding "shared" causes the code to run successfully.
Comment #2 by debio264 — 2010-11-06T18:38:34Z
Created attachment 804
Patch to correct the issue
Comment #3 by spam — 2010-11-06T18:51:56Z
this was already fixed in the latest release of dmd2.050
Comment #4 by debio264 — 2010-11-06T18:54:09Z
(In reply to comment #3)
> this was already fixed in the latest release of dmd2.050
Well, this is from an install of dmd2.050 on Windows, so something didn't get updated properly.
Comment #5 by debio264 — 2010-11-06T18:57:27Z
(In reply to comment #4)
> (In reply to comment #3)
> > this was already fixed in the latest release of dmd2.050
>
> Well, this is from an install of dmd2.050 on Windows, so something didn't get
> updated properly.
My apologies, I seem to have 2.049. Guess I installed the new version on the other machine I develop on.
Comment #6 by schveiguy — 2010-11-08T05:18:37Z
Properly marking this. Bugs that are valid on some version of D that is not the latest version are not invalid, they can be duplicates or resolved (could have been a bug that was resolved but had no matching bug in bugzilla).
Thanks
*** This issue has been marked as a duplicate of issue 4344 ***