RE: Bug found in SDPAnnounceParser

Hi Andreas and others,

I had a similar issue some time ago with an even nastier device sending
mixed "\r" and "\n" EOLs in the same SDP. At one point the original
algorithm (Revision 1.7  2006/07/13 09:02:37) would find a '\r' in the
current line and match it with a '\n' in the very last line, ignoring
everything else in between. That device was a really big mess but we had
no way to change its behavior.

So I decided to write a more strict implementation for the
SDPAnnounceParser constructor that looks a lot like the one that you
wrote but covering these nasty cases:

public SDPAnnounceParser(String message) 
{
  int start = 0;
  String line = null;
  // Return trivially if there is no sdp announce message
  // to be parsed. Bruno Konik noticed this bug.
  if (message == null ) return;
  sdpMessage = new Vector();
  // Strip off leading and trailing junk.
  String sdpAnnounce = message.trim() + "\r\n";
  // Bug fix by Andreas Bystrom.

// ** my changes start here ** 

  while (start < sdpAnnounce.length())
  {
    int lfPos = sdpAnnounce.indexOf("\n", start);
    int crPos = sdpAnnounce.indexOf("\r", start);

    if (lfPos > 0 && crPos < 0)
    {
      // there are only "\n" separators
      line = sdpAnnounce.substring(start, lfPos);
      start = lfPos + 1;
    }
    else if (lfPos < 0 && crPos > 0)
    {
      //bug fix: there are only "\r" separators
      line = sdpAnnounce.substring(start, crPos);
      start = crPos + 1;
    }
    else if (lfPos > 0 && crPos > 0)
    {
      // there are "\r\n" or "\n\r" (if exists) separators
      if (lfPos > crPos)
      {
        // assume "\r\n" for now
        line = sdpAnnounce.substring(start, crPos);
        // Check if the "\r" and "\n" are close together
        if (lfPos == crPos + 1)
          start = lfPos + 1; // "\r\n"
        else
          start = crPos + 1; // "\r" followed by the next record and a
"\n" further away
      }
      else
      {
        // assume "\n\r" for now
        line = sdpAnnounce.substring(start, lfPos);
        // Check if the "\n" and "\r" are close together
        if (crPos == lfPos + 1)
          start = crPos + 1; // "\n\r"
        else
          start = lfPos + 1; // "\n" followed by the next record and a
"\r" further away
      }
    }
    else if (lfPos < 0 && crPos < 0) // end
      break;

// ** my changes end here ** 

    sdpMessage.addElement(line);
  }
}

Would this cover you case as well? 

Regards,

Ricardo