(In reply to comment #5)
> did this work for you?
>
Yes it did.

did this work for you?

Jean-Marie,

Please give revision 8142 or later a try and see if it works for you. There are even some unit tests in there to test for NaN and nan, now.

• r8142 /opennms/trunk/opennms-dao/src/ (2 files in 2 dirs):
  Improve error checking on print values in getPrintValue and add some more tests.
  See .

$ rrdtool create /tmp/test.rrd DS:test:GAUGE:1000:U:U RRA:AVERAGE:test && rrdtool graph /dev/null DEF:ds=/tmp/test.rrd:test:AVERAGE PRINT:ds:AVERAGE:%le
0x0
nan

jmk points out:
rrdfetch man page says that : <<UNKNOWN data representation depends on the OS's printf function>>

What is the value that can't be parsed by Double.parseDouble? Also, I'm not sure if using NaN if we hit any exceptions with Double.parseDouble is a good idea... it might hide errors where something other than a value equivalent to NaN is spit out by rrdtool and friends.

I did a quick test with bean shell (bsh), and it was able to happily parse NaN:

bsh % System.out.println(Double.parseDouble("NaN"));
NaN
bsh %