Wrong indentation after one line functions ending with a bitstring

[larshesel]

defmodule X do
  def a, do: <<1 :: size(8)>>
      def b, do: <<2 :: size(8)>>
          def c, do: <<3 :: size(8)>>
end

[mattdeboard]

So I tried adding a regression test for this but oddly enough, it passed, but it caused another test to fail.

;; Will pass when #180 is resolved.
(elixir-def-indentation-test indent-after-bitstring/1
                             (:expected-result :failed :tags '(indentation))
"
defmodule X do
  def a, do: <<1 :: size(8)>>
      def b, do: <<2 :: size(8)>>
          def c, do: <<3 :: size(8)>>
end"
"
defmodule X do
  def a, do: <<1 :: size(8)>>
      def b, do: <<2 :: size(8)>>
          def c, do: <<3 :: size(8)>>
end")

This passed (unexpectedly), but had this for output:

Ran 70 tests in 0.276 seconds
7 expected failures
2 unexpected results:
   PASSED  elixir-ert-indent-after-bitstring/1
   FAILED  elixir-ert-indent-multiline-match
Test suite failed!

Obviously the latter test normally passes. How weird. There must be some kind of state mutation that we're not resetting after tests...

[tonini]

@mattdeboard @larshesel I suppose I found the issue.

Because of the bitstring, elixir-smie-backward-token it's checking for elixir-smie--binary-sequence-regexp and returns OP.

So on line 3 below:

   defmodule X do
     def b, do: <<1 :: size(8)>>
->   def b, do: <<2 :: size(8)>>
     def c, do: <<3 :: size(8)>>
   end

Following rules get called twice:

:before 'do:'; sibling-p:nil parent:((155) 18 def) prev-is-OP:nil hanging:nil == nil
:after 'OP'; sibling-p:nil parent:(114 25 do:) prev-is-OP:nil hanging:nil == (column . 2)

With the following changes to the :after 'OP' rule it would be fixed.

(`(:after . "OP")
 (cond
  ((smie-rule-sibling-p) nil)
  ((smie-rule-hanging-p) (smie-rule-parent elixir-smie-indent-basic))
  ((and (not (smie-rule-sibling-p))
        (not (smie-rule-hanging-p))
        (smie-rule-parent-p "do:"))
   (smie-rule-parent (- (+ elixir-smie-indent-basic elixir-smie-indent-basic))))
  (t (smie-rule-parent))))

[tonini]

Obviously the latter test normally passes. How weird. There must be some kind of state mutation that we're not resetting after tests...

@mattdeboard yeah we should have a look at this. At the moment a lot of tests failing for me locally, I tried it with the master, so it's really confusing.

[larshesel]

@mattdeboard I just tested out the add-test-case-180 branch and the issue looks solved to me 👍. haven't run the tests, though.