Inflater.c bug?

pra at mint.se pra at mint.se
Tue Jan 26 03:28:39 PST 1999


Hi,
I have problems using GZIPInputStream. I always get an error message
like this:
java.lang.Error: oversubscribed dynamic bit lengths tree
        at java/util/zip/InflaterInputStream.read(72)
        at java/util/zip/GZIPInputStream.read(78)
        at java/io/InputStreamReader.read(53)
        at java/io/BufferedReader.refillBuffer(164)
        at java/io/BufferedReader.readLine(130)
        at TestIo.main(line unknown, pc 0x83e812d)

I have tracked the error down to inftrees.c in zlib but can't find any
way of improving the code. Any one having any experinec on this.

I have tested it with
kaffe 1.0b2 and kaffe 1.0b3 and the latest snapshot. It has also been
tested both agains zlib 1.0.4 and 1.1.2.

This is sad becaus I get a 7 timespeed improvement in io with kaffe
compared to the blackdown JDK, and I need this improvement for GZ-files
to.

Here is the codefragment creating the input-stream, from wich I 
read with readLine():

  protected Reader makeReader(String file) throws java.io.IOException {
    Reader reader;
    // Create an input source for the lexer
    FileInputStream fileStream = new java.io.FileInputStream( file );
    if ( se.dn.annons.util.FileUtils.isGZ( file) ) {
      GZIPInputStream inStream = new GZIPInputStream( fileStream );
      reader = new BufferedReader( new InputStreamReader( inStream ) );
    } else {
      reader = new BufferedReader( new InputStreamReader( fileStream));
    }
    return reader;
  }

Greetings
Peter
-- 
--------------------------------------------------------
Peter Antman		AB Dagens Nyheter
System Developer	Gjörwellsgatan 30, 105 15 Sockholm
Email: pra at mint.se	WWW: http://www.dn.se/
Phone: 08-738 1338	Mobile: 070-675 3942 
--------------------------------------------------------



More information about the kaffe mailing list