RFR JDK-8167461: jshell tool: Scanner#next() hangs tool

[Robert Field]

Looks good.  Might be safest to copy the checks from the top of 
InputStream's implementation --

     public int read(byte b[], int off, int len) throws IOException {
         if (b == null) {
             throw new NullPointerException();
         } else if (off < 0 || len < 0 || len > b.length - off) {
             throw new IndexOutOfBoundsException();
         } else if (len == 0) {
             return 0;
         }

I'm somewhat unclear on why the read/write in Util are changed to the 
buffered form.  BTW.  There isn't (and wasn't) an EOF test here.

As to input carrying over: I can see either as being surprising in some 
cases.  As-is, very simple "Enter two numbers: " kinds of newbie coding 
will misbehave, and misbehave in a way that could look like a hang 
(waiting on input).
With a input carry-over implementation however, garbage entered, perhaps 
much earlier in the session would still be what is read later -- that 
could be hard to diagnose, and would take coding (or a /reset) to cure.
I think addressing that should be a separate issue.

-Robert

On 10/12/16 12:41, Jan Lahoda wrote:
> Hello,
>
> Bug:
> https://bugs.openjdk.java.net/browse/JDK-8167461
>
> The problem is that this sequence of snippets is not working properly 
> (won't read the System.in properly/seemingly hangs):
> jshell> Scanner input = new Scanner(System.in);
> jshell> int x = input.nextInt();
>
> The reason is that the PipeInputStream (which is used to pass data 
> received by the remote agent from the socket to System.in) does not 
> override the read(byte[], int, int) method. So the default 
> implementation will be used, which waits until all the requested data 
> are read. Scanner wrap s the System.in with InputStreamReader, and 
> that appears to read (up to) 8192 bytes into a buffer, but the user 
> typically does not type so many characters. The solution is to 
> override the read(byte[], int, int) method, and return only bytes we 
> have buffered (except the method will wait for the first byte).
>
> Webrev:
> http://cr.openjdk.java.net/~jlahoda/8167461/webrev.00/
>
> Also, more general question related to the System.in handling: unused 
> input for one snippet is not passed to following snippets - is that 
> OK, or do we need to discuss changing that? I.e.:
> jshell> int x = input.nextInt();
> 1 1
> x ==> 1
> //the other "1" is thrown away
>
> jshell> int x = input.nextInt();
> 2
> x ==> 2
> //the user had to type additional input.
>
> Thanks,
>     Jan