Expression parsed differently in v1.12.5 (again)

[schillic]

There was a parser change in v1.12.0 that has been fixed (#59911) and seems to have been finally backported to v1.12.5 (but still not yet to v1.13.0-beta2 by the way). However, things still do not work fully as expected with a larger example, which is given below (and used to work before v1.12).

# v1.11.9 (expected result):
julia> :(x' = A * x)
:(x' = A * x)

# v1.12.4:
julia> :(x' = A * x)
:(x' = begin
          #= REPL[1]:1 =#
          A * x
      end)

# v1.12.5:
julia> :(x' = A * x)
:(x' = (#= REPL[1]:1 =#)(*, A, x))

# v1.13.0-beta2 (only for the record; same as v1.12.4):
julia> :(x' = A * x)
:(x' = begin
          #= REPL[1]:1 =#
          A * x
      end)

As the OP in #59911, I would mainly like to know whether this is the new intended behavior or considered an unintended change that is going to be changed back.

[mlechu]

Should be changed to the no-block version, just hasn't landed in 1.13 yet. It looks like the fix landed in 1.12.5 (#60612) and will be in 1.13 when #60920 is merged. I can't find any differences based on the expression's right-hand side on either version, but do let me know if you've found any.

[topolarity]

I believe the issue is about how the :call now contains a LineNumberNode in 1.12.5:

julia> dump(:(x' = A * x))
Expr
  head: Symbol =
  args: Array{Any}((2,))
    1: Expr
      head: Symbol '
      args: Array{Any}((1,))
        1: Symbol x
    2: Expr
      head: Symbol call
      args: Array{Any}((4,))
        1: LineNumberNode
          line: Int64 1
          file: Symbol REPL[1]
        2: Symbol *
        3: Symbol A
        4: Symbol x

[mlechu]

My bad, I think I've conditioned myself to tune LineNumberNodes out. Fix up at #61062

[topolarity]

Thanks for the helpful bug report @schillic !

This should be fixed in the next 1.12 / 1.13 patch releases

[schillic]

Wonderful - thanks for the prompt fix!