RFR 9: 8155808: Process.getInputStream().skip() method is faulty

[Roger Riggs]

Hi Pavel,

On 6/17/2016 5:03 AM, Pavel Rappo wrote:
> Hi Roger,
>
> Do we need PipeInputStream to be a java.io.FileInputStream? Can we get away with
> it being an InputStream? If so, then we could use a simple delegation which
> would allow us to avoid code duplication (the implementation of `skip` is almost
> identical to one in InputStream):
Subclassing is the normal (one) way to override/augment behavior of an 
existing class.
And I was looking for a minimal change since there have been many issues 
in this area.
Also, I avoided the extra object and performance overhead of delegation 
by extending FIS.

Roger
>
>      private static class PipeInputStream extends InputStream {
>
>          private final FileInputStream fis;
>
>          PipeInputStream(FileDescriptor fd) {
>              fis = new FileInputStream(fd);
>          }
>
>          @Override
>          public int read() throws IOException {
>              return fis.read();
>          }
>
>          @Override
>          public int read(byte[] b) throws IOException {
>              return fis.read(b);
>          }
>
>          @Override
>          public int read(byte[] b, int off, int len) throws IOException {
>              return fis.read(b, off, len);
>          }
>
>          @Override
>          public void close() throws IOException {
>              fis.close();
>          }
>
>          @Override
>          public int available() throws IOException {
>              return fis.available();
>          }
>      }
>
> Thanks,
> -Pavel
>
>> On 13 Jun 2016, at 16:06, Roger Riggs <Roger.Riggs at oracle.com> wrote:
>>
>> A latent issue with Process InputStreams support for the skip method was reported[1].
>>
>> Though the InputStream returned is a BufferedInputStream, in some cases,
>> the skip method calls FileInputStream.seek on the pipe containing output
>> from the process but the pipes do not support seek.  The proposed fix is to override
>> the skip method to implement skip as reading the bytes instead of invoking seek.
>>
>> Please review and comment:
>>
>> Webrev:
>>
>> http://cr.openjdk.java.net/~rriggs/webrev-skip-8155808/index.html
>>
>> Issue:
>>
>>   [1] https://bugs.openjdk.java.net/browse/JDK-8155808
>>
>> Thanks, Roger
>>
>>