Re: [cp-patches] Re: FYI: java/io/FileInputStream.java& java/io/FileOutputStream.java

[Dalibor Topic]

Jeroen Frijters wrote:
> Dalibor Topic wrote:
>
> > Yeah, i can understand the inefficient part, as the same code is 
> > duplicated in two classes. Could you elaborate some more on 
> > the racy part?
>
>
> If another application creates a directory after the check, but before
> the file is created, an incorrect error is thrown. So ideally the native
> code that calls the platform file open API needs to handle this error
> case.

Isn't that the case in any way, independently of the implementation 
strategy used? Unless a platform has an atomic 'check if dir and open' 
function call, there is always a time slice where some other application 
can do stuff that invalidates what we know.

Another strategy to implement this in the FileChannelImpl constructor 
would be to call open() first, and then to check whether what we have 
opened is a directory, and if it is one to call close() + throw 
FileNotFoundException in that case. Does that sound OK?

I can attach a patch, if you're interested.

cheers,
dalibor topic