Expectations of empty-command-newline #3 are unmeetable (was: Tests XPASS on master)

[danielshahaf]

make test is currently failing due to an XPASS:

# empty-command-newline
1..4
ok 1 - [1,1] «:»
ok 2 - [2,2] «;»
ok 3 - [3,3] «↵» - # TODO "issue #616"
ok 4 - cardinality check

We should change the test file so it either is expected to pass and passes, or is expected to fail and fails.

@phy1729 Were you working on this? What's the correct fix here?

[phy1729]

That test has been failing since added a few weeks ago. I assumed you were working in the area.

Now that I look at it, shouldn't the third point be '3 3 commandseparator "issue #616"' # \n rather than unknown-token (i.e. expect the correct result)?

[danielshahaf]

That test has been failing since added a few weeks ago. I assumed you were working in the area.

I'm not. I wasn't aware that travis was failing, though.

Now that I look at it, shouldn't the third point be ` '3 3 commandseparator "issue #616"' # \n` rather than `unknown-token` (i.e. expect the correct result)?

I'm not sure what it should highlight. I suppose it shouldn't be unknown-token because the sequence semicolon-newline is muscle memory for C and Perl programmers (the rationale behind highlighting BUFFER=': ; ;' as an error doesn't apply here). As to commandseparator, I wonder: should ';' and $'\n' command separators be highlighted differently?

[danielshahaf]

I've added a duct tape fix to the "CI is red" problem and am repurposing this issue to track implementing a proper fix. See the commit message of ab4b6f5 for details.