Not detecting _FillValue of scaled data correctly

[guygriffiths]

This comes from a bug report for EDAL (Reading-eScience-Centre/edal-java#109).

In the test data available at that bug report, the fill values are not being detected properly. The variable tmp is stored as an integer and has a _FillValue of 99999 and a scale_factor of 1.e-05. By the time we get to the isMissing_() method in ucar.nc2.dataset.EnhanceScaleMissingImpl, the value has been scaled to 0.99999. However, when that method calls isFillValue() on that, it returns false, since it is testing against the unscaled value (i.e. 99999).

[cwardgar]

Fixed in #1067. Thanks for the report!

[guygriffiths]

Perfect, thanks. Would you be able to cut a 5.0.0-alpha4 release including this? We're currently depending on 5.0.0-alpha3 and it'd be nice to have this change.

[lesserwhirls]

Hi @guygriffiths - I'll cut a second beta release later today or tomorrow and this change will be in there :-)

[lesserwhirls]

@guygriffiths - sorry for the delay on our end! 5.0.0-beta2 has the fix and is now available as a release on our nexus server.