is_disabler: do not flatten positional args

[xclaesse]

No description provided.

[dcbaker]

This is a behavior change that is absolutely going to break someone's build.

And of course, there isn't an obvious way (in my mind) to fix this because of the way disablers are handled.

[xclaesse]

It's actually not a breaking change because is_disabler([foo, bar]) already fails because it accepts only 1 positional argument.

[xclaesse]

The use-case here is something like that:

list = [foo, bar]
if something
  list = disabler()
endif
...
if is_disabler(list) # This used to fail because list is flatten into `is_disabler(foo, bar)` and that func allows a single argument
  ...
endif

[xclaesse]

Note there is a subtle workaround: is_disabler(list.lenght()), because any method call on a disabler does return a disabler.

[dcbaker]

Oh, well if it already fails because the list gets flattened, then that seems fine. I retract my objections.

Some docs would be nice, but otherwise seems fine.

[xclaesse]

Added doc for the behavior with a list argument. Also added a FeatureNew.

[eli-schwartz]

It appears this change breaks a test that explicitly assumes summary() can take a dictionary with elements that are disablers.

[xclaesse]

It appears this change breaks a test that explicitly assumes summary() can take a dictionary with elements that are disablers.

Hm, I fear this change could have too many unforeseen consequences like that. I changed this PR to not change the behavior of is_disabled() (can be revisited later) and make the public function is_disabler() reflects whatever is_disabled() does with lists/dicts. My initial goal here was really to make them both consistent.

[xclaesse]

I intentionally left is_disabler() doc not mention what happens with dictionaries to not close the door to future changes.