Variable scope bug?

[Andrew Todd]

I've got a problem with the scope of a variable, consider
the following example:


MyCOMObject.MyO bjectClass obj;

try
{
obj = new MyCOMObject.MyO bjectClass();
returnXml = obj.getEnquiryU sers();
}
catch (Exception e)
{
//handle error
}
finally
{
if (obj!=null)
{
Marshal.Release ComObject(obj);
obj = null;
}
}

This won't compile, it tells me that the variable obj is
unassigned when I test if it's null in the finally
statement.

This doesn't seem correct to me; I need to release the
COM object in the finally statement as if there is an
error in the try block the object will never be released.
These steps work in Java and are the steps we normally
take when interfacing with COM objects, is this a bug?

Thanks
Andrew

[Stu Smith]

Re: Variable scope bug?

If the statement "new MyCOMObject.MyO bjectClass()" were to throw an
exception then obj would not have been assigned, and thus the warning in the
finally clause.

You need to decide whether exceptions resulting from creating the object
need to be handled here (in which case set obj to null beforehand), or
whether you aren't handling them here (in which case create the object
outside the try block).

HTH,

Stu


"Andrew Todd" <andrew.todd@uk online.co.uk> wrote in message
news:0bbc01c3a9 de$fe1552b0$a40 1280a@phx.gbl.. .[color=blue]
> I've got a problem with the scope of a variable, consider
> the following example:
>
>
> MyCOMObject.MyO bjectClass obj;
>
> try
> {
> obj = new MyCOMObject.MyO bjectClass();
> returnXml = obj.getEnquiryU sers();
> }
> catch (Exception e)
> {
> //handle error
> }
> finally
> {
> if (obj!=null)
> {
> Marshal.Release ComObject(obj);
> obj = null;
> }
> }
>
> This won't compile, it tells me that the variable obj is
> unassigned when I test if it's null in the finally
> statement.
>
> This doesn't seem correct to me; I need to release the
> COM object in the finally statement as if there is an
> error in the try block the object will never be released.
> These steps work in Java and are the steps we normally
> take when interfacing with COM objects, is this a bug?
>
> Thanks
> Andrew
>[/color]

[Jon Skeet [C# MVP]]

Re: Variable scope bug?

Andrew Todd <andrew.todd@uk online.co.uk> wrote:[color=blue]
> I've got a problem with the scope of a variable, consider
> the following example:[/color]

<snip>
[color=blue]
> This won't compile, it tells me that the variable obj is
> unassigned when I test if it's null in the finally
> statement.[/color]

Indeed, and it's right. Set it to null to start with. Local variables
don't have any default value; they need to be definitely assigned
before they're read.
[color=blue]
> This doesn't seem correct to me; I need to release the
> COM object in the finally statement as if there is an
> error in the try block the object will never be released.
> These steps work in Java and are the steps we normally
> take when interfacing with COM objects, is this a bug?[/color]

No, it wouldn't work in Java. Here's an example showing it:

public class Test
{
public static void main(String[] args)
{
String x;
try
{
x = "hello";
}
catch (Exception e)
{
System.out.prin tln (e);
}
finally
{
if (x==null)
{
System.out.prin tln ("Oops");
}
}
}
}

And compiling it:

Test.java:16: variable x might not have been initialized
if (x==null)
^
1 error

--
Jon Skeet - <skeet@pobox.co m>

http://www.pobox.com/~skeet

If replying to the group, please do not mail me too

[Andrew Todd]

Re: Variable scope bug?

[color=blue]
>Indeed, and it's right. Set it to null to start with.[/color]
[color=blue]
>No, it wouldn't work in Java. Here's an example showing[/color]
it:

Of course - you're right, I'd forgotten to set it to null
initially!

Thanks
Andrew