Re: BufferedOutputStream patch

[jmc@cmpharm.ucsf.edu]

OK, here's what the JDK behavior is... it does in fact differ
between BufferedWriter and BufferedOutputStream, but not between
JDK 1.1 and 1.2 (tested on 1.1.6 on Irix, and 1.2pre1 on Linux):

OS, byte - flush on overflow
OS, byte[] - overflow
Wr, char - overflow
Wr, char[] - fill(!)

And here's the current libjava behavior:

OS, byte - overflow
OS, byte[] - fill
Wr, char - fill
Wr, char[] - fill

The docs you quoted in the previous msg imply there is a difference
between Writer and OutputStream behavior, although not between single
character and array operations.

My patch "fixes" the first difference (in BOS) to bring it in line
with the JDK; the difference in BWr should be patched also, IMO.

My opinion is that in cases where the JDK docs are unclear on
inconsistent, libjava should duplicate observed JDK behavior... I
assume comparing testcase ouput is kosher as long as one doesn't
actually look at the JDK source.  In cases where the JDK behavior
is contrary to the documentation, that's a different problem!
Sun is pretty good about fixing one or the other...

For anyone who wants to duplicate my testcase for Writers, here's
a code fragment; note that one has to be careful since OutputStreamWriter
also buffers its output.  This buffer size cannot be specified, so
one has to flush to make sure nothing is there:

  OutputStreamWriter osw = new OutputStreamWriter(System.out);
  Writer outfile = new BufferedWriter(osw,6);
  for (int i=0; i<6; i++)
      outfile.write('X');
  // or:  2 writes of a char[] containing XXX
  osw.flush();

JMC
-- 
John-Marc Chandonia (jmc@cmpharm.ucsf.edu)              We're everywhere...
Cohen Lab, University of California San Francisco       for your convenience.
http://yuri.harvard.edu/~jmc                                -- Psi Corps <*>