variable attributes showing up in OPeNDAP global attributes!

[rsignell-usgs]

This is pretty crazy, but if you take this netcdf file and put it on THREDDS (I've tried the most recent version 4.6.13 (latest) and 4.6.11 (circa 2017)), the variable attributes from the variable called error show up in the global attributes on OPeNDAP (as well as in the variable attributes)!

It's pretty goofy to have _FillValue and long_name showing up in the global attributes!

https://rsignell-test.whoi.edu/thredds/dodsC/silt/usgs/Projects/stellwagen/CF-1.6new/MATANZAS_18/11102sigs_burst-cal.nc.html

[DennisHeimbigner]

Can you post the .dds and .das that opendap generates for this file?

[rsignell-usgs]

I'm guessing you mean more than putting the .das and .dds on the end of the URL, right?
https://rsignell-test.whoi.edu/thredds/dodsC/silt/usgs/Projects/stellwagen/CF-1.6new/MATANZAS_18/11102sigs_burst-cal.nc.das
https://rsignell-test.whoi.edu/thredds/dodsC/silt/usgs/Projects/stellwagen/CF-1.6new/MATANZAS_18/11102sigs_burst-cal.nc.dds

[DennisHeimbigner]

Actually no, I just meant what you sent, thanks. My hypothesis is that
the problem is occurring in the translation from .nc to .dds and .das.
So I thought I would start by looking at those.

[DennisHeimbigner]

When I look at that url with the netcdf-c ncdump, it looks ok to me.
What are you using to view it? ToolsUI?

[rsignell-usgs]

I'm just looking at
https://rsignell-test.whoi.edu/thredds/dodsC/silt/usgs/Projects/stellwagen/CF-1.6new/MATANZAS_18/11102sigs_burst-cal.nc.html

[rsignell-usgs]

The problem was caused by the variable called error in the NetCDF file. If we rename that to something else, the problem goes away.

[DennisHeimbigner]

I was looking at that. The opendap code has a number of reserved names that
it uses to pass non-standard information. Apparently "error" is one one of those,
but I had had no luck finding it yet.

[rsignell-usgs]

https://www.unidata.ucar.edu/mailing_lists/archives/thredds/2016/msg00180.html