Workaround for #190 and #207

For some reason, when Package Control updates Anaconda in Windows, the
winsocks enter in an inconsistent state that produce several problems,
being the most critical one a total unresponsiveness of the ST3.

What Package Control do is just disable the package, install the update
and then re-enable de package, if you do that manually, it seems to work
just fine. I am being able to reproduce this behaviour after anaconda
automatic updates only.

The commit just detect when a socket enter in inconsistent state, then
show an error message to the user and stop the asynchronous poll (this
should not make ST3 unresponsive).

anaconda_lib/ioloop.py

@@ -274,24 +274,39 @@ def loop():
     """Main event loop
     """
 
+    def restart_poll(error):
+        print('Unhandled exception in poll, restarting the poll request')
+        print(error)
+        for traceback_line in traceback.format_exc().splitlines():
+            print(traceback_line)
+
+        with IOHandlers()._lock:
+            for handler in IOHandlers()._handler_pool.values():
+                handler.close()
+            IOHandlers()._handler_pool = {}
+
     def inner_loop():
 
         while NOT_TERMINATE:
             try:
                 poll()
                 time.sleep(0.01)
+            except OSError as error:
+                if os.name != 'posix' and error.errno == os.errno.WSAENOTSOCK:
+                    msg = (
+                        'Unfortunately, the Windows socket is in inconsistent'
+                        ' state, restart your sublime text 3. If the problem '
+                        'persist, fill an issue report on:'
+                        '   https://github.com/DamnWidget/anaconda/issues'
+                    )
+                    print(msg)
+                    import sublime
+                    sublime.error_message(msg)
+                    terminate()
+                else:
+                    restart_poll(error)
             except Exception as error:
-                print(
-                    'Unhandled exception in poll, restarting the poll request'
-                )
-                print(error)
-                for traceback_line in traceback.format_exc().splitlines():
-                    print(traceback_line)
-
-                with IOHandlers()._lock:
-                    for handler in IOHandlers()._handler_pool.values():
-                        handler.close()
-                    IOHandlers()._handler_pool = {}
+                restart_poll(error)
 
         # cleanup
         for handler in IOHandlers()._handler_pool.values():