|
Why would two threads call dispose on the same object ?
Christian Graus
Please read this if you don't understand the answer I've given you
"also I don't think "TranslateOneToTwoBillion OneHundredAndFortySevenMillion FourHundredAndEightyThreeThousand SixHundredAndFortySeven()" is a very good choice for a function name" - SpacixOne ( offering help to someone who really needed it ) ( spaces added for the benefit of people running at < 1280x1024 )
|
|
|
|
|
Thanks Christian,
My concern is, if a reference is or potentially held by two threads, and they may call Dispose at the same time. Any comments?
regards,
George
|
|
|
|
|
They may not. Strictly speaking, if two threads or objects hold references to a single object, neither of them may call Dispose.
If they do, they are violating the fundamental rules of the IDisposable pattern.
The rules in short:
1. Dispose is meant to tell an object this: "I now know for sure that nobody has a reference to you except for myself, and I will never talk to you again - you may now safely throw away any resources you've been holding on to. If I or someone else ever talks to you again, please throw an exception.".
2. If you don't know for sure that you're the only one holding a reference to the object, you should never call Dispose. Just drop your reference (as in myRef = null;). The GC (garbage collector) will then call its finalizer (~TheClass) as soon as a) it realizes that nobody holds a reference to the object any more and b) it or your program has nothing more important to do. The finalizer will call Dispose(false), ensuring that any unmanaged (operating system handles and whatnot) will finally be released.
The only reason that the docs say that you should always allow calls to Dispose even if you are already disposed or not completely initialized is that you may in some cases get called from a catch or finally block that shouldn't have to check whether or not the try block that should have called you actually did it before the exception got thrown. This can never occur simultaneously on different threads, though.
So, the IDisposable stuff is only an optimization, enabling us to force expensive resources to be released before the GC gets around to it - but it should never be used unless we know for sure that the object is dead for good. Note that there is a lot of code (even in the framework) that just crashes spectacularly instead of throwing the proper exceptions if you try to use an object that has been disposed. Try loading a bitmap, showing it in a picture box and then disposing it (while still showing it). Results will vary - the app may exit silently, throw a protection fault or even display random graphics. At least, this was true in 1.1 - it was a while since I tested that.
Later,
--
Peter
|
|
|
|
|
Great reply, Peter!
1.
PeterTheSwede wrote: The only reason that the docs say that you should always allow calls to Dispose even if you are already disposed or not completely initialized is that you may in some cases get called from a catch or finally block that shouldn't have to check whether or not the try block that should have called you actually did it before the exception got thrown. This can never occur simultaneously on different threads, though.
I am very interested in the scenario you described, but still confused after reading a couple of times. Could you show some pseudo code please?
2.
"dead for good" means on one else refers the object?
3.
PeterTheSwede wrote: Note that there is a lot of code (even in the framework) that just crashes spectacularly instead of throwing the proper exceptions if you try to use an object that has been disposed.
You mean ObjectDisposedException should be thrown -- and in your points should be the correct behavior -- in this scenario?
regards,
George
|
|
|
|
|
1. I'll try.
If I remember the recommendations for the IDisposable interface correctly, the requirement that duplicate (but not simultaneous, mind you) calls to Dispose() shouldn't fail is explained with something like this:
MyClass obj = new MyClass();
try
{
obj.DoSomethingThatCanFail();
obj.Dispose();
this.DoSomeOtherStuff();
}
catch
{
obj.Dispose();
throw;
}
This code should be considered correct - that is, you shouldn't have to check if things went wrong at A or B in the catch block. Note that obj could potentially be a 2GB image, and the DoSomeOtherStuff() method could also be something extremely memory-hungry. In which case it makes sense to call obj.Dispose() then and there instead of in a finally block.
2. Exactly.
No one but you refers to it and you aren't ever going to use it again. If you can't guarantee those conditions, you're essentially not allowed to Dispose() the object.
3. Yes.
According to Microsofts current recommendations, ObjectDisposedException should be thrown if you try to do anything with an object that is disposed (except calling Dispose() again). But like I said, there's a lot of code out there that just crashes instead. Understandably - it's easy to forget if (disposed) throw new ObjectDisposedException(); in a method or property somewhere.
Which makes it even more important to never use Dispose() unless you're absolutely sure that the object is dead. People who get buried alive don't play nice... 
Later,
--
Peter
|
|
|
|
|
Thanks Peter,
Your reply is great!
1.
In your sample, it is possble to call Dispose twice on the same object if there is exception thrown from DoSomeOtherStuff, in this case, there will be ObjectDisposedException from catch block?
2.
Is it always a program bug when we met with ObjectDisposedException? Means someone Disposed the object instance, but some other guys may still refer it and call the method on it.
3.
I like your below code,
if (disposed) throw new ObjectDisposedException();
Currently, I think of two solutions when met with ObjectDisposedException. Solution 1 is to do operations on object straigtforward without checking IsDisposed property, and wrap catch block to catch ObjectDisposedException, solution 2 is to check IsDisposed property as you showed above, and it is set, do nothing but throw ObjectDisposedException to propogate out or write a log without rethrow. I am wondering which pros and cons and your suggestions of which solution is better?
regards,
George
|
|
|
|
|
Thanks for the appreciation!
When English isn't your own language (which I assume it isn't for neither of us), it can be hard to understand even the good C#/CLR books sometimes. I have read a lot of good texts about the IDisposable interface over the last few years (there are even some excellent articles on CodeProject), but I must admit that it wasn't until recently that all the pieces of the puzzle fell into place. Sometimes you have to read the same thing (phrased differently) several times before you get it.
So, I'll try to summarize the most important points again (it should answer your follow-up questions as well), as short and concise as I can:
A: It is always legal to call Dispose() on an object more than once, for exactly one reason: The example I gave (and similar code) should always work even if something happens in DoSomeOtherStuff(). This means that a call to Dispose() should never cause an ObjectDisposedException - if the object is already disposed, Dispose() should just silently do nothing. Note that usually, the only place where this is acceptable is in an exception handler (as in my example).
B: It is always illegal to call any other method or property on an object after Dispose() has been called. And ideally, an ObjectDisposedException should always be thrown if that happens (but like I said, this is not always the case - sometimes very bad things such as unmanaged memory corruption or similar happen instead - even in some .NET Framework classes).
Some conclusions (some obvious, some not):
C: You may only call Dispose() on an object if a) you created the object, b) you haven't given away the reference to anyone and c) you're done with it. Can't be said too many times...
D: If you ever see an ObjectDisposedException, it is a bug. It should be fixed (yesterday), so it's fine to just crash (as long as you leave a clue as to why, meaning that if you're a service or a web application, you should log the exception before crashing - if you're a Windows app, you should display it and/or log it).
E: A repetition of rule C from a different angle: If you ever expose your reference to a disposable object (by returning it from a property or using it in a call into a class you don't control), you effectively give up your right to dispose of it. There are cases when it is OK to do differently, but they must be documented. For example: "If you dispose of a MyParent object, all MyChild objects created by it will also be disposed of, and may not be used again" (if I remember correctly, there are such relationships in the SqlClient namespace in the framework). In such cases it does make sense to deviate a bit from the basic rules.
F: Answering your original question - making the Dispose() method thread-safe is meaningless. It should only be called by the class that created the object, only when it will not be used again (even though additional Dispose() calls may be made by the same caller due to exception handlers) and only when nobody else can hold a reference to the object. Hence, the class creating the object has no reason whatsoever to call Dispose() from more than one thread. If someone comes up with a reason (I once did, actually - I called the same termination and cleanup routine from a worker thread that I exposed to clients when it died due to external problems), the caller is free to implement his own synchronization (I did in my termination routine).
G: A public IsDisposed property is entirely meaningless. If the code calling Dispose() doesn't know it did, it's broken (the example with the exception handler is the only exception - but since an extra Dispose() should cost less than a property check, the property still doesn't make sense).
H: The IDisposable interface is meant for classes holding unmanaged resources (including other disposable objects). Don't bother otherwise (unless you're writing an abstract class that would typically be implemented using unmanaged resources or disposable objects).
Finally, an example from the framework: If you create an Image from a BinaryStream you may think that you can dispose of the BinaryStream when that is done. Not so. The Image will not read the entire stream until it has to, and if the stream is a file it is even free to read the same part of the file several times when needed. Look at rule E. You showed the stream to the image, so you're no longer in control of it, and may not call Dispose(). In this particular case, you can dispose of the image (after making sure that nothing that has seen it is still around) and then of the stream.
Clear as mud?
One more suggestions: Do read the articles on implementing the IDisposable interface here on CodeProject as well. There are some aspects (such as how exceptions in Dispose() should be handled, and how object inheritance comes into play) to it that I haven't mentioned. It's too complex for a forum post, and I don't remember all of it anyway...
Good Night!
--
Peter
|
|
|
|
|
Thanks Peter,
Your reply is so great!
After reading your reply, I think it should be a bug -- either my code or CLR code, if ObjectDisposedException is thrown.
My detailed scenario is like this,
(I am writing an Http Server service using HttpListner class of C#, and handles coming request in an asynchronous way.)
1. The exception from server side, when I call EndGetContext, not from client side;
2. On the server side, I create an HttpListener object, and waiting for incoming client request;
3. When I received a request, I will process the request asynchronously by using BeginGetContext, and in this way HttpListener object instance can still listen for incoming calls, while processing backend at the same time;
4. When sending the request to server from client, the client will close the connection immediately -- during the asynchronous processing period;
5. The server will throw exception when I call EndGetContext. At the same time, ObjectDisposedException will happen.
Any ideas why the exception will happen? I am also interested in whose Disposed method is called, any ways to detect which object instance's Disposed method is invoked?
regards,
George
|
|
|
|
|
Hi!
Again, thanks for the appreciation. I usually don't write such long answers but I've been at a customer site (staying at a hotel) this week, and nights get a bit boring...
Sorry, no idea what is happening. I have never used asynchronous calls seriously so I don't have any experience to draw from. Then again, try looking for asynchronous samples (I don't find any using HttpListeners, but there are bound to be others) and understand what they do. Then perhaps it will become clearer.
--
Peter
|
|
|
|
|
Hi,
I've seen others have this concern and have seen a multitude of solutions, but like Christian says, why? I only realized this recently (and felt pretty stupid about that), but like he says, there is NO circumstance whatsoever where it is correct for two different threads to invoke Dispose() on the same object.
In fact, if Dispose() is ever invoked twice on the same object (even from the same thread), it is a design fault. It is explicitly wrong to invoke Dispose() on an object if there is a chance (no matter how slim) that someone else is holding a reference to it (meaning that if you ever directly or indirectly expose a reference to an object outside of your class, you should never dispose the object).
Further, Dispose() cannot be invoked unless the object is alive, so it can never conflict with the finalizer being invoked by the GC.
The most common pattern for implementing IDisposable that I've seen is:
protected virtual void Dispose(bool disposing)
{
if (!disposed)
{
if (disposing)
{
}
disposed = true;
}
}
public void Dispose()
{
Dispose(true);
GC.SuppressFinalize(this);
}
~MyClass()
{
Dispose(false);
}
Note: Written from memory, beware of spelling mistakes. Google for "IDisposable pattern" and you will get all the examples you need (good and bad).
Later,
--
Peter
|
|
|
|
|
|
I was wondering if there was any way to see what functions are availble in dll's. For example how can I find out the available functions for the user32.dll?
|
|
|
|
|
www.pinvoke.net has a good list.
Christian Graus
Please read this if you don't understand the answer I've given you
"also I don't think "TranslateOneToTwoBillion OneHundredAndFortySevenMillion FourHundredAndEightyThreeThousand SixHundredAndFortySeven()" is a very good choice for a function name" - SpacixOne ( offering help to someone who really needed it ) ( spaces added for the benefit of people running at < 1280x1024 )
|
|
|
|
|
|
If you really want to get oldschool, use the dumpbin[^] utility.
Scott P
“It is practically impossible to teach good programming to students that have had a prior exposure to BASIC: as potential programmers they are mentally mutilated beyond hope of regeneration.”
-Edsger Dijkstra
|
|
|
|
|
Thanks for alll your replies.
|
|
|
|
|
I have to include a image such that it is editable when clicked on it.
Can any one help me?
|
|
|
|
|
|
http://msdn.microsoft.com/en-us/magazine/cc164043.aspx[^]
1) Open up your form in design view
2) In the toolbox, Right click and select "Choose Items"
3) When the dialog pops up, click the "Com components" tab
4) scroll down to "Microsoft Visio x.x Drawing Control" and tick it.
5) Hit OK
6) The control will appear in the toolbox under General.
7) Drag the control onto your form.
You can use this control to display a Visio drawing and allow the user to edit it. See the MSDN article above for a tutorial of using the control.
Note that this will bind to a specific version of Visio. The user must have that version installed to use your app. There is a way of bundling up a lite installer with your app, that the article talks about. I would assume the user still has to have a Visio license, but I'm not sure.
Simon
|
|
|
|
|
hi,i have a form and when you click a button you show a webbrowser whith a flash and a textbox in the html,and i would obtain the information of the textbox(is in the web page) how i can?
|
|
|
|
|
You need to do this by inspecting the website using the DOM. It's not a simple thing, the web browser is it's own world, it doesn't expose stuff like a windows app
Christian Graus
Please read this if you don't understand the answer I've given you
"also I don't think "TranslateOneToTwoBillion OneHundredAndFortySevenMillion FourHundredAndEightyThreeThousand SixHundredAndFortySeven()" is a very good choice for a function name" - SpacixOne ( offering help to someone who really needed it ) ( spaces added for the benefit of people running at < 1280x1024 )
|
|
|
|
|
Hi,
I'm using an Access file as data source of a crystal report Report, in my computer all is ok, but when install my application in another system, because the path of mdb file changes, the report don't work?
how can give a related path in crystal report or force it to read the path from a setting file?
Best wishes
|
|
|
|
|
mehrdadc48 wrote: I'm using an Access file as data source of a crystal report Report, in my computer all is ok, but when install my application in another system, because the path of mdb file changes, the report don't work?
how can give a related path in crystal report or force it to read the path from a setting file?
You will need to define the connection parameters to the report within your code.
HTH.
|
|
|
|
|
hi
i need a control that display my image maps and panning and zooming it in my control.
i was create a userControl named ImageMap, and add a panel to it and set Doc to Fill and wrote this code to get result :
public partial class ImageMap : UserControl<br />
{ <br />
Bitmap bitmap; <br />
BufferedGraphicsContext currentContext;<br />
BufferedGraphics myBuffer; <br />
PointF viewPortCenter;<br />
float Zoom = 1.0f;<br />
<br />
bool draging = false;<br />
Point lastMouse;<br />
public static Rectangle rec;<br />
<br />
<br />
public ImageMap()<br />
{<br />
InitializeComponent();<br />
currentContext = BufferedGraphicsManager.Current;<br />
setup(false);<br />
rec = this.panel1.DisplayRectangle;<br />
} <br />
<br />
private void setup(bool resetViewport)<br />
{ <br />
if (myBuffer != null)<br />
myBuffer.Dispose();<br />
myBuffer = currentContext.Allocate(this.panel1.CreateGraphics(), this.panel1.DisplayRectangle);<br />
if (bitmap != null)<br />
{<br />
if (resetViewport)<br />
SetViewPort(new RectangleF(0, 0, bitmap.Width, bitmap.Height)); <br />
} <br />
this.panel1.Focus();<br />
this.panel1.Invalidate();<br />
} <br />
<br />
private void SetViewPort(RectangleF worldCords)<br />
{ <br />
if (worldCords.Height > worldCords.Width)<br />
{<br />
this.Zoom = worldCords.Width / bitmap.Width;<br />
}<br />
else<br />
this.Zoom = worldCords.Height / bitmap.Height;<br />
<br />
viewPortCenter = new PointF(worldCords.X +(worldCords.Width / 2.0f), worldCords.Y + (worldCords.Height / 2.0f));<br />
<br />
} <br />
<br />
private void PaintImage()<br />
{<br />
if (bitmap != null)<br />
{<br />
float widthZoomed = panel1.Width / Zoom;<br />
float heigthZoomed = panel1.Height / Zoom;<br />
<br />
if (widthZoomed > 30000.0f)<br />
{<br />
Zoom = panel1.Width / 30000.0f;<br />
widthZoomed = 30000.0f;<br />
}<br />
if (heigthZoomed > 30000.0f)<br />
{<br />
Zoom = panel1.Height / 30000.0f;<br />
heigthZoomed = 30000.0f;<br />
}<br />
<br />
if (widthZoomed < 2.0f)<br />
{<br />
Zoom = panel1.Width / 2.0f;<br />
widthZoomed = 2.0f;<br />
}<br />
if (heigthZoomed < 2.0f)<br />
{<br />
Zoom = panel1.Height / 2.0f;<br />
heigthZoomed = 2.0f;<br />
}<br />
<br />
float wz2 = widthZoomed / 2.0f;<br />
float hz2 = heigthZoomed / 2.0f;<br />
Rectangle drawRect = new Rectangle(<br />
(int)(viewPortCenter.X - wz2),<br />
(int)(viewPortCenter.Y - hz2),<br />
(int)(widthZoomed),<br />
(int)(heigthZoomed));<br />
<br />
<br />
myBuffer.Graphics.Clear(Color.White);
<br />
myBuffer.Graphics.DrawImage(bitmap, this.panel1.DisplayRectangle, drawRect, GraphicsUnit.Pixel);<br />
myBuffer.Render(this.panel1.CreateGraphics());<br />
} <br />
}<br />
<br />
protected override void OnBackgroundImageChanged(EventArgs e)<br />
{<br />
bitmap = (Bitmap)this.BackgroundImage;<br />
setup(true);<br />
base.OnBackgroundImageChanged(e);<br />
}<br />
<br />
protected override void OnResize(EventArgs e)<br />
{<br />
setup(false);<br />
base.OnResize(e);<br />
}<br />
<br />
private void panel1_Paint(object sender, PaintEventArgs e)<br />
{ <br />
PaintImage();<br />
}<br />
<br />
private void panel1_MouseWheel(object sender, System.Windows.Forms.MouseEventArgs e)<br />
{<br />
Zoom += Zoom * (e.Delta / 1200.0f);
if (e.Delta > 0)
viewPortCenter = new PointF(viewPortCenter.X + ((e.X - (panel1.Width / 2)) /(2* Zoom)), viewPortCenter.Y + ((e.Y - (panel1.Height/2)) / (2*Zoom))); <br />
this.panel1.Invalidate(); <br />
}<br />
<br />
private void panel1_MouseDown(object sender, MouseEventArgs e)<br />
{<br />
if (e.Button == MouseButtons.Left)<br />
draging = true;<br />
}<br />
<br />
private void panel1_MouseMove(object sender, MouseEventArgs e)<br />
{<br />
if (draging)<br />
{<br />
viewPortCenter = new PointF(viewPortCenter.X + ((lastMouse.X - e.X)/Zoom), viewPortCenter.Y + ((lastMouse.Y- e.Y)/Zoom)); <br />
panel1.Invalidate(); <br />
}<br />
lastMouse = e.Location;<br />
}<br />
<br />
private void panel1_MouseUp(object sender, MouseEventArgs e)<br />
{<br />
if (e.Button == MouseButtons.Left)<br />
draging = false;<br />
} <br />
}
but in runTime, when i panning image in my control, it has a flicker and very bad result, for solve this problem, in top of above code, i define a new class that derived from Panel and wrote this code :
public class overRidePanel : Panel<br />
{<br />
protected override void OnPaintBackground(PaintEventArgs pevent) <br />
{<br />
<br />
}<br />
}
then modify my panel to instanciate from new overRidePanel class, as u can see in above code, i was override OnPaintBackground event only, then run my app, it gave me best result and remove flicker when i panning and zomming my images in app, but when my app run, the ImageMap does not have a any image to display it, and my control in runTime does not display white color correctly(it show back of the my app and not good for me), i want when i start app, it is white color before selecting image, but how to do ?
thanks
|
|
|
|
|
Hello,
I've been trying for while now to upload a picture and get it to be displayed in an Image control. I still can't get the code right. I used
FileUpload1.PostedFile.SaveAs(filepath);
Image1.ImageURL = filepath;
after performing the necessary checks. The ASP.NET debugger says the path format is not supported. I need help.
Thanks.
Obinna from Nigeria.
PS. I'm using VS 2005.
//obinnaaj Eof
|
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
