Patch: FYI: fix no-progress InputStreamReader bug

[Tom Tromey <tromey at redhat dot com>]

I'm checking this in on the trunk.  Most likely I will put it in 4.0
once the branch reopens.

We noticed that ecj was hanging on some sources when given '-encoding
utf8'.  There are two bugs here.

First, 'utf8' was getting the iconv converter and not our built-in
UTF-8 converter.  I added a new alias for this.

Second, we had a regression in InputStreamReader.  There is a
situation in the current code where InputStreamReader will make no
progress and thus will hang.  The fix is to notice this and play a
mark/reset game so we can refill the buffer.  This is a hack, but
hopefully all this code will be going away once the nio stuff is done.

I added a new mauve test case for this failure.

Tom

Index: ChangeLog
from  Tom Tromey  <tromey@redhat.com>
	* java/io/InputStreamReader.java (refill): Handle no-progress
	case correctly.
	* gnu/gcj/convert/IOConverter.java: Add 'utf8' alias.

Index: java/io/InputStreamReader.java
===================================================================
RCS file: /cvs/gcc/gcc/libjava/java/io/InputStreamReader.java,v
retrieving revision 1.20
diff -u -r1.20 InputStreamReader.java
--- java/io/InputStreamReader.java 30 Jun 2005 03:19:49 -0000 1.20
+++ java/io/InputStreamReader.java 6 Jul 2005 20:03:20 -0000
@@ -289,9 +289,23 @@
 	  return -1;
 	converter.setInput(in.buf, in.pos, in.count);
 	int count = converter.read(buf, offset, length);
-	in.skip(converter.inpos - in.pos);
-	if (count > 0)
-	  return count;
+
+	// We might have bytes but not have made any progress.  In
+	// this case we try to refill.  If refilling fails, we assume
+	// we have a malformed character at the end of the stream.
+	if (count == 0 && converter.inpos == in.pos)
+	  {
+	    in.mark(in.count);
+	    if (! in.refill ())
+	      throw new CharConversionException ();
+	    in.reset();
+	  }
+	else
+	  {
+	    in.skip(converter.inpos - in.pos);
+	    if (count > 0)
+	      return count;
+	  }
       }
   }
 }
Index: gnu/gcj/convert/IOConverter.java
===================================================================
RCS file: /cvs/gcc/gcc/libjava/gnu/gcj/convert/IOConverter.java,v
retrieving revision 1.7
diff -u -r1.7 IOConverter.java
--- gnu/gcj/convert/IOConverter.java 20 Mar 2004 00:24:49 -0000 1.7
+++ gnu/gcj/convert/IOConverter.java 6 Jul 2005 20:03:20 -0000
@@ -1,4 +1,4 @@
-/* Copyright (C) 2000, 2001  Free Software Foundation
+/* Copyright (C) 2000, 2001, 2005  Free Software Foundation
 
    This file is part of libgcj.
 
@@ -28,6 +28,8 @@
     // canonical name.
     hash.put ("iso-latin-1", "8859_1");
     hash.put ("iso8859_1", "8859_1");
+    // At least one build script out there uses 'utf8'.
+    hash.put ("utf8", "UTF8");
     // On Solaris the default encoding, as returned by nl_langinfo(),
     // is `646' (aka ASCII), but the Solaris iconv_open() doesn't
     // understand that.  We work around the problem by adding an