msmetadata.scansforfield sometimes reports spurious error

[pkgw]

Reported by Thoughtworks folks (CC @pandearti).

On OSX, sometimes msmetadata.scansforfield throws an error that doesn't make much sense. Here's a screenshot. In short, msmetadata_cmpt.cc::scansforfield throws an exception saying

Request field ID 0 is greater than or equal to the number of records (4294967295) in this MS's FIELD table.

(This is apparently casa::MSMetaData::_hasFieldID() in casacore/ms/MSOper/MSMetaData.cc line 4346.) It sounds like the problem doesn't happen for all data sets, but does happen consistently; but I'm not sure about that.

But the fieldnames method works.

This was seen on OSX 10.10 and 10.11, Python 2.7, casa-nrao-casacore=4.5, casa-python=4.6.0, casa-tools=4.5.

[pkgw]

OK, same question as in issue #3: your email said that you had switched to using the conda-forge and pkgw-forge sets of packages rather than the outdated pkgw packages. But the version information in this bug report seems to refer to the old versions. Can you please reply with the versions that you're testing now? The latest available ones should be something like casacore = 2.2.0_1, casa-tools = 4.7.2_1, and casa-python = 4.7.2_np112py27_4, where the piece after the underscore is the "build number" reported by the conda list command.

[pandearti]

Here are the new package versions (as displayed by conda list command)
Yes we still have same issue with these new packages.

casa-data r11000 0 pkgw-forge
casa-python 4.7.2 np112py27_3 pkgw-forge
casa-tools 4.7.2 1 pkgw-forge
casacore 2.2.0 0 conda-forge