Hi Yurong,
Thanks for reporting the case: it is indeed a bug. I have fixed it and committed the change to the code base of 3DNA v2.0. Before I have the time to produce an updated binary of v2.0 for users to try out, you can simply ignore it, or set H-bonding distance cut-off to 3.5 A instead of 4.0 A with the "-p" option. I will let you know when I have a new compiled version for Linux/Mac OS X/Cygwin etc.
FYI, the bug was due to a very special case where a character string was not initialized. The program produced that garbage string where it should have been empty.
Keep in mind that, to me, the more bugs you report, the merrier! 3DNA becomes more robust by fixing bugs ...
Xiang-Jun