Make PR#48 actually work

[jefshe]

So it turns out my last pull request didn't fix the issue and the unit test I added fails locally.

Sorry about that, do the unit tests not get run on travis?

[eddelbuettel]

Strange. The unit tests run under R CMD check. it should also have come up at your end if you did R CMD build ... followed by R CMD check ....

[eddelbuettel]

And it does fail locally here:

[...]
✔  checking for unstated dependencies in ‘tests’
─  checking tests ...  Running ‘runUnitTests.R’
    ERROR
   Running the tests in ‘tests/runUnitTests.R’ failed.
   Last 13 lines of output:
     Error in attributes(xobj) <- attrib[names(attrib) != "class"] : 
       'names' attribute [11] must be the same length as the vector [1]
     test.serialize.sublist: (1 checks) ... OK (0 seconds)
     test.serialize_pb: (2 checks) ... OK (0.01 seconds)
     test.serialize_pb.alldatasets: (1 checks) ... OK (0.17 seconds)
     > 
     > ## Return success or failure to R CMD CHECK
     > if (getErrors(tests)$nFail > 0) {
     +    stop("TEST FAILED!")
     + }
     > if (getErrors(tests)$nErr > 0) {
     +    stop("TEST HAD ERRORS!")
     + }
     Error: TEST HAD ERRORS!
     Execution halted
✔  checking for unstated dependencies in vignettes
✔  checking package vignettes in ‘inst/doc’
[...]

(using Gabor's rcmdcheck from the littler script rcc.r I use for this).

And it passes with your patch. Very strange that it did not trip Travis CI.

(We had some recent 'red' there I upgraded the underlying script to use R 3.5.* by defaul but the external repo I use for the protocol buffer libraries still ties us to R 3.4.*. I will change that here by switching to testing in Docker as I have done in a few other repos requiring an external and not-so-common library.)