Re: [Patch] Fix PR libgcj/20389

[David Daney <ddaney at avtrex dot com>]

Bryce McKinlay wrote:
> Hi David,
>
> It looks like this bug may already be fixed in GNU Classpath:
>
> 2004-12-07  Jeroen Frijters  <jeroen@frijters.net>
>
>        * java/io/BufferedInputStream.java
>        (skip): Don't return -1 on EOF.
>
> 2004-12-06  Jeroen Frijters  <jeroen@frijters.net>
>
>        * java/io/BufferedInputStream.java
>        (read(byte,int,int)): Handle case when zero bytes are requested.
>
> 2004-07-10  Jeroen Frijters  <jeroen@frijters.net>
>
>        * java/io/BufferedInputStream.java (count, pos, marklimit): Removed
>        unnecessary initialization. (marktarget, CHUNKSIZE): Removed.  
>        (bufferSize): New field. (close): Reset state. (read()): Removed
>        unnecessary cast. (read(byte[],int,int): Try harder to read the
>        requested number of bytes. (mark,reset,refill): Fixed handling of
>        marking. (skip): Added stream closed check and removed mark 
> handling.
>
>
> Since this class ought to be merged, could you try using 
> BufferedInputStream.java from Classpath and see if that fixes it?

BufferedInputStream.java from Classpath breaks InputStreamReader.java.

InputStreamReader.java from Classpath breaks because it needs 
gnu.java.io.decode.Decoder.

We either need to merge much of Classpath, or fix the existing 
BufferedInputStream.

The Classpath version looks plausible, but I would have to hack on it to 
make it fit with InputStreamReader.

Your call I guess.

David Daney