Bug in java.io.InputStreamReader

[Anthony Green]

Hi - thanks for your bug report.  A few comments....

> I have egcs-2.93.19 and libgcj (I don't know how to determine the
> version number, but the most recent entry in Changelog is dated
> 1999-04-26).

ChangeLog dates are as good as anything right now.

> I've included a small test program, a Makefile, and two minimally
> different sample inputs (100 and 101 bytes long).  ''

This is useful - thank you.

Are you familiar with Mauve? See http://sourceware.cygnus.com/mauve .
In a perfect world, all test case submissions would fit right into
this testing framework.  This is what we are using to test our
libraries, as are the Kaffe and Classpath people.  Admittedly, there's
not a whole lot of documentation on how to write proper test cases,
and there are plenty of bad examples to go by :-)

> bin: $(CLASSES)
> 	LD_RUN_PATH=$(LIBGCJ) \
> 	gcj -g -o $(EXECUTABLE) -fmain=$(MAIN) -L$(LIBGCJ) $(CLASSES)

If you've installed libgcj where you've installed egcs, you shouldn't
have to use -L.

Also, recent development snapshots of automake understand java source
and class files, and will do the right thing.  See
http://sourceware.cygnus.com/automake .

AG

-- 
Anthony Green                                               Cygnus Solutions
                                                       Sunnyvale, California