Batch EJH Pull requests

[WardF]

• Fixes nc_test/tst_open_cdf5.c does not include config.h #842

Creating a pull request combining all outstanding ejh pull requests, in a branch with the .wif suffix. This will reduce the amount of CI tests which occur. The merge looks great, except for a few compile errors in Visual Studio. Addressing those and then will get things merged on in, clearing out the ejh backlog.

[WardF]

@edhartnett So I'm able to duplicate the bug reported in #843 working with this group of combined pull requests, and have confirmed that the issue described is absent in the master branch. Before I separate this pull request back into the component parts, perhaps you have an idea off the top of your head which pull request this stems from? I can then remove it from this batch and merge everything but the problematic one, until we have a fix for it. Thanks for your help with this!

[edhartnett]

Stand by, I will take a look and get back to you...

[edhartnett]

@WardF I am building NCO so I can run this. Running into some problems I am working out with Charlie offline. I will come back when I have an answer.

[WardF]

Sounds good, if there are any broader netCDF issues that should be considered, you and @czender should feel free to pull myself and/or @dmh in. Thanks, I'll sit on this so that we're primed to merge once this is complete.

[WardF]

Ok, since the issue isn't related to your pull requests, I'll go ahead and merge all these shortly. Sorry for the confusion. I went back to the master version I tested on yesterday and it doesn't have this error but also says it's up to date, so I assume something is out of sync there. I blew the directory away and started with a fresh pull, now.

[edhartnett]

Awesome.

Incidentally there is another LATEFILL issue open (not new behavior), one of the tests (tst_vars2.c) hits an error but instead of erroring out, just prints an error message. So it is ignored. But it is a LATEFILL error that (I believe) Quincey expected to occur, but apparently does not. I wrote an issue but have not otherwise touched it: #117.

According to @wkliao this should now be fixed: so the test should be changed to error out now instead of just printing an error message, and the #117 can be closed. I can do that as a separate PR or one of you guys can roll it in. That would be an extra test for correct behavior.