You're awesome! Thanks =)
The original code of #4772 with end formats and compiles with this fix.
However the workaround using {{ "end".id }} still produces an error in the formatter (but compiles fine).

Sorry, I forgot to fix the formatter.

Done fixing the formatter.

Confirmed, now it works as expected. Thanks again!

Can we get this approved?

I think I'd prefer a parse_macro_control_body method, instead of repeating incr, parse_body, decr.

What is nest for? Any nested macro content?

nest is for counting if, while and other usual (not macro) control block nesting to detect macro block's end. But it behaves like stack against macro control internally, in other codes:

macro foo
  # `nest == 0`
  {% if true %}
    # `nest == 0` also because `{% if true %}` is not counted by `nest`
    if true
      # `nest == 1` because `if true` is counted
  {% end %}

  # Now, `nest == 0`. oops! after closing `{% if true %} ... {% end %}`, `nest` is reset before it.
  
  {% if true %}
    # `nest == 0` and the parser found `end`, so the parser mistakes it as macro body `end`, and then #4769 bug is caused.
    end
  {% end %}
end

Macro body parsing is very very and very complex task (hmm, @asterite?), so unfortunately we cannot count macro/usual control block nesting in simple way.

Of course, to correct this behavior is the right way to fix #4769, but I think this PR way is also right and this is simpler than that. The most preferable way is to apply both, but I think this makes it hard to review.

Yes, this is a fix for incorrect behaviour and does not add much complexity. It makes sense to merge this as a bugfix for now and address refactoring in a different PR.

Thanks the detailed explanation @makenowjust. That makes sense.