Problem with negative lookahead + left recursion

[fsh]

Using ~x + ... seems to disallow whitespace skip with left recursion. Is this intended behavior? If so, what's the idiomatic way of doing it in left recursive grammars?

An example:

import pyparsing as pp
pp.ParserElement.inline_literals_using(pp.Suppress)
pp.ParserElement.enable_left_recursion()

Top = pp.Forward()

# an xyz word that doesn't start with xxx
Atom = ~pp.Literal('xxx') + pp.Word('xyz')
#          ^--------- this negative lookahead seems 
#                     to mess with whitespace handling
#                     in recursive grammars?
#
# This workaround works:
# Atom = pp.White()[0,1] + ~pp.Literal('xxx') + pp.Word('xyz')

# (Minibug: pp.White(min=0) cannot match 0 whitespace)

Expr0 = pp.Forward()
Expr0 <<= Expr0 + '[' + Top + ']' | Atom

# ... in my case there's many levels of expressions,
# and they're constructed dynamically ...

Expr1 = Top
Expr1 <<= Top + '*' + Expr0 | Expr0

The thought is that this would match anything like x[ x * x[y] ][y] * y with optional whitespaces anywhere, but whitespace is only skipped inside expressions, not at the start:

>>> Top.run_tests(['x[y*x]', 'x [y * xx ]', '  x', 'x[ y*xx]', '  x*y'])

x[y*x]
['x', 'y', 'x']

x [y * xx ]
['x', 'y', 'xx']

  x
['x']

x[ y*xx]
x[ y*xx]
 ^
ParseException: Expected end of text, found '['  (at char 1), (line:1, col:2)
FAIL: Expected end of text, found '['  (at char 1), (line:1, col:2)

  x*y
  x*y
   ^
ParseException: Expected end of text, found '*'  (at char 3), (line:1, col:4)
FAIL: Expected end of text, found '*'  (at char 3), (line:1, col:4)