for loops could be supported with a compiler fix

[disruptek]

Here's the test that runs against the almost branch:

https://github.com/disruptek/cps/blob/almost/tests/fors.nim

import testes
import cps

type
  C = ref object of RootObj
    fn*: proc(c: C): C {.nimcall.}

proc trampoline(c: C) =
  var c = c
  while c != nil and c.fn != nil:
    c = c.fn(c)

testes:

  block:
    proc loop() {.cps: C.} =
      for i in 0 .. 3:
        discard
    trampoline loop()

Here's the broken rewrite; CPS doesn't know what to do with the var i statement. I don't even understand how it passes through sem successfully to reach our macro...

 16  proc foo() =
 17    block :tmp:
 18      var i
 19      ## An alias for `countup(a, b, 1)`.
 20           ## 
 21           ## See also:
 22           ## * [..<](#..<.i,T,T)
 23      var res = 0
 24      block :tmp:
 25        while res <= 3:
 26          i = T(res)
 27          discard
 28          inc(res, 1)

[alaviss]

#84 has the walkaround

[disruptek]

The AST is so very awful that we cannot use getImplTransformed and so this effort is being killed in favor of a more manual transformation. 💀

[zevv]

We could always file this as a CPS issue on nim and pass it to Araq

[disruptek]

He would tell us that we made our bed when we chose to do this outside the compiler, and now we get to lie in it. And he's not wrong.