Bug in java.nio.StringCharBuffer.slice

[Alan Bateman]

John Platts wrote:
> Here is the current implementation of slice in 
> java.nio.StringCharBuffer (as found in OpenJDK 7):
> 44     public CharBuffer slice() {
> 45         return new StringCharBuffer(str,
> 46                                     -1,
> 47                                     0,
> 48                                     this.remaining(),
> 49                                     this.remaining(),
> 50                                     this.position());
> 51     }
>
> Here is the implementation of slice used for heap-based byte buffers, 
> which is found in java/nio/Heap-X-Buffer.java.template:
>  97     public $Type$Buffer slice() {
>  98         return new Heap$Type$Buffer$RW$(hb,
>  99                                         -1,
> 100                                         0,
> 101                                         this.remaining(),
> 102                                         this.remaining(),
> 103                                         this.position() + offset);
> 104     }
>
> It appears that the slice() method fails to return the correct region 
> of the string for CharBuffer objects that wrap a string whenever 
> offset is non-zero. Here is the corrected code, with the correction in 
> red:
> 44     public CharBuffer slice() {
> 45         return new StringCharBuffer(str,
> 46                                     -1,
> 47                                     0,
> 48                                     this.remaining(),
> 49                                     this.remaining(),
> 50                                     offset + this.position());
> 51     }
>
> Is this bug limited to OpenJDK 7, or is this bug found in OpenJDK 6, 
> JDK 6, JDK 5, and JDK 1.4? This bug might be affecting code that uses 
> CharBuffers.
Thanks for the bug report. I've created a bug to track it:
7000913: (bf) CharBuffer.wrap, slice, position, slice leads to 
CharBuffer with incorrect offset

I'm kinda surprised this hasn't been noticed before but it may be that 
it's rare to slice a buffer that itself was created by slicing a wrapped 
char sequence. It does highlight a hole in our tests so part of fixing 
this will adding further tests.

So are you contributing the fix to this? I see a John Platts on the SCA 
list but I'm not sure if that is you.

As regards versions, then it is likely to exist in all versions.

-Alan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openjdk.java.net/pipermail/core-libs-dev/attachments/20101117/27f055bc/attachment.html>