Robert Snyder Robert Snyder - 1 month ago 10
C# Question

ObjectDisposedException: Safe handle has been closed

So this is a rather small question with a big explanation. As is noted by the title I am getting an unhandled exception telling me my Safe handle has been closed. What I'll probably have to do is edit this post a few times with more and more code to help me diagnose what the problem is.

I'm using POS for .NET to make a Service Object for my RFID and MSR device. Although my devices are the same, I have 2 different Virtual COM Port chips that communicate to those devices. One by Silicon labs, the other by FTDI. I wanted to use the plug and play features with POS for .NET so I gave it both my Hardware ID's. Because it is plug and play I have the full hardware path available to me which I can then create a SafeFileHandle using a call to PInvoke and using that SafeFileHandle I create a FileStream. The FTDI chip doesn't let me talk to the devices directly like that so I have to get the friendly name of the device then use mutex to pull out the COM port then create a SerialPort instance. That step works fine and great. As a FYI I have tried to use the Friendly name of both chips to get the COM port and the Silicon Labs one (for some strange reason) doesn't get listed using SetupAPI.GetDeviceDetails using the Ports GUID. I'm not sure on that one since in Device Manager the Silicon labs Device Class Guid is the Ports GUID.

Well since both the SerialPort and the FileStream have a Stream object I decided to use that to read and write to that port. The problem with that is if I send a RFID command to the MSR device the MSR device doesn't respond back with anything. So if I use this code

int fromReader = ReaderStream.ReadByte();
my thread is blocked. It's a blocking call and requires a minimum of 1 byte to proceed. So I looked around and it appears the only solution is to use a separate thread and set a timeout. If the timeout happens then abort the thread.

Thread t = new Thread(new ThreadStart(ReadFromStream));
t.Start();
if (!t.Join(timeout))
{
t.Abort();
}


(t.Abort has been surrounded with a try/catch to no avail, since it didn't fix the problem I removed it)

ReadFromStream is Abstract method in RFID Device. Here is one of the implementations

protected override void ReadFromStream()
{
var commandLength = USN3170Constants.MIN_RESPONSE_LENGTH;
var response = new System.Collections.Generic.List<byte>(USN3170Constants.MIN_RESPONSE_LENGTH);
for (int i = 0; i <= commandLength; i++)
{
int fromReader = ReaderStream.ReadByte();
if (fromReader == -1) break; //at end of stream
response.Add((byte)fromReader);

if (response.Count > USN3170Constants.DATA_LENGTH_INDEX && response[USN3170Constants.DATA_LENGTH_INDEX] > 0)
{
commandLength = response[USN3170Constants.DATA_LENGTH_INDEX] + 3;
}
}

streamBuffer = response.ToArray();
}


(
int fromReader = ReaderStream.ReadByte();
was surrounded with a try/catch. Only thing it caught was the aborted thread exception, so I took it out)

The above code is where I suspect the problem lies. The strange thing is, though, is that I have a unit test which I feel mimics rather well the Microsoft Test App.

(FYI QUADPORT is the FTDI chipset)

PosExplorer posExplorer;
DeviceCollection smartCardRWs;
[Test]
public void TestQuadPortOpen()
{
posExplorer = new PosExplorer();
smartCardRWs = posExplorer.GetDevices(DeviceType.SmartCardRW, DeviceCompatibilities.CompatibilityLevel1);
//if using quadport one item is the MSR and the other is the RFID
//because of that one of them will fail. Currently the first Device in the collection is the the RFID, and the second is MSR
Assert.GreaterOrEqual(smartCardRWs.Count, 2);
//Hardware Id: QUADPORT\QUAD_SERIAL_INTERFACE
foreach(DeviceInfo item in smartCardRWs)
{
Assert.AreEqual("QUADPORT\\QUAD_SERIAL_INTERFACE", item.HardwareId);
}

SmartCardRW rfidDevice = (SmartCardRW)posExplorer.CreateInstance(smartCardRWs[0]);
SmartCardRW msrDevice = (SmartCardRW)posExplorer.CreateInstance(smartCardRWs[1]);

rfidDevice.Open();
Assert.AreNotEqual(ControlState.Closed, rfidDevice.State);
rfidDevice.Close();

try
{
msrDevice.Open();
Assert.Fail("MSR Device is not a RFID Device");
}
catch
{
Assert.AreEqual(ControlState.Closed, msrDevice.State);
}

rfidDevice = null;
msrDevice = null;
}


When I run that test I do not get the SafeFileHandle exception. In fact the test passes.

So I am at a loss as to how to track down this bug. Since I'll be using this Service Object in a different program that I am also creating I'll probably end up using this code from this test in that program. However I feel that the Microsoft Test App is more or less the "Golden Standard". Is it really... probably not. But it does work good for my purposes, SO I feel it is a problem with my code and not theirs.

Any tricks on how I can narrow this down? FYI I've tried using the debugger but walking the Open Code the error does not occur. I also walked the Update Status Timer and it also does not throw the error. Once I hit continue then I'll get the exception. I turned of Just My Code and Loaded Symbols and it tells me "Source Information is missing from teh debug information for this module"

Answer

This problem (and in particular the reference to a SerialPort instance) sounds suspiciously like the problem documented at http://connect.microsoft.com/VisualStudio/feedback/details/140018/serialport-crashes-after-disconnect-of-usb-com-port.

As I understand it, in the case of a non-permanent SerialPort (like one associated with a USB device, for example) when the port "goes away" unexpectedly the underlying Stream associated with it gets disposed. If there is an active read or write operation on the port at the time a subsequent call to SerialPort.Close can lead to the exception you mention, however the exception is occurring in Microsoft's code running on a different thread and cannot be caught from within your code. (It will still be seen by any "last chance" exception handler you have bound to the UnhandledException event on the AppDomain.)

There seem to be two basic workaround styles in the linked document. In both instances, after opening the port you store a reference to the BaseStream instance for the open port. One workaround then suppresses garbage collection on that base stream. The other explicitly calls Close on the base stream, capturing any exceptions thrown during that operation, before calling Close on the SerialPort.

EDIT: For what it's worth, under the .NET framework V4.5, it appears that none of the documented workarounds on the Microsoft Connect site fully resolve the problem although they may be reducing the frequency with which it occurs. :-(

Comments