regression(1.2=>1.4 onwards): enum with same name in different scope gives wrong RT results

[timotheecour]

Example

when true:
  # proc main = # ditto
  template main =
    block:
      type A = enum a0, a1, a2
      for ai in A:
        echo (ai,)
    block:
      type A = enum b0, b1, b2, b3
      for ai in A:
        echo (ai,)

  static: main() # ok
  main() # bug

Current Output

vm:
(a0,)
(a1,)
(a2,)
(b0,)
(b1,)
(b2,)
(b3,)

rt:
(a0,)
(a1,)
(a2,)
(a0,)
(a1,)
(a2,)
(3 (invalid data!),)

Expected Output

rt should be like vm:
(a0,)
(a1,)
(a2,)
(b0,)
(b1,)
(b2,)
(b3,)

Additional Information

• works 1.2.8
• fails 1.4.0
• fails 1.5.1 74a8f23

Note that it works in vm or in js but not in c backend
EDIT: seems very similar to #15526
(and also to #5170 but that was fixed)

[timotheecour]

after git bisect the culprit is 2629d61 #15091 /cc @Clyybber:

git bisect start v1.4.2 v1.2.0
git bisect run zsh -c "source ~/.zshrc && nim=$nim_csources_local_X d_nim_build_simple_local --skipusercfg --hints:off -o:bin/nim.bisect && bin/nim.bisect c -r --skipparentcfg --skipusercfg -d:case4 $timn_D/tests/nim/all/t11895.nim"