Finally as statement raises exception in next try/except block

[gradha]

The following example shows an undesired behaviour of finally as a statement:

proc do_stuff(out_filename: string): bool =
  var out_file: TFile
  try:
    if not out_file.open(out_filename, fmWrite):
      echo "Could not open ", out_filename, " for writing."
      return
  except EOutOfMemory:
    return

  finally: out_file.close

  try: out_file.write("yeah\n")
  except EIO:
    echo "Could not write header"
    return

  result = true

when isMainModule:
  echo "Did stuff? ", do_stuff("test.txt")

When run the output is

Could not write header
Did stuff? false

The file test.txt is created but empty. If the finally line is commented out, the example works and the test file gets written to. This suggest that finally forces the next try/except block to fail.

[skyfex]

Simplified test case for this bug:

try:
  discard
except EOS:
  discard

finally: 
  echo "B"

try: 
  echo "A"
except EIO:
  discard

A should come before B

[skyfex]

Oh, duh.. Nimrod interprets the finally block to be part of the try-except structure right above it.

So in a sense, it's not a bug? Question is if a newline between the except and finally blocks should indicated that the finally statement should not be part of the above try-except structure.

The language is a bit ambiguous here. How can we resolve that?

[Araq]

Empty lines don't have a meaning currently. It's a language wart, not a bug. I'm closing this but feel free to start a discussion on the forum about it. My vote is to do nothing about it.