Bug fix: octal syntax is $'\377', not $'\0377'.

This is inconsistent with echo -e '\0377'.

(Related to issue #61: Alpine uses $'\n'.)

core/id_kind.py

@@ -399,7 +399,10 @@ def _AddKinds(spec):

  # For C-escaped strings.
  spec.AddKind('Char', [
      'OneChar', 'Stop', 'Hex', 'Octal', 'Unicode4', 'Unicode8', 'Literals'
      'OneChar', 'Stop', 'Hex',
      # Two variants of Octal: \377, and \0377.
      'Octal3', 'Octal4',
      'Unicode4', 'Unicode8', 'Literals'
  ])

core/word_compile.py

@@ -42,7 +42,16 @@ def EvalCStringToken(id_, value):
  elif id_ == Id.Char_Stop:  # \c returns a special sentinel
    return None

  elif id_ == Id.Char_Octal:
  elif id_ == Id.Char_Octal3:  # $'\377'
    s = value[1:]
    i = int(s, 8)
    if i >= 256:
      i = i % 256
      # NOTE: This is for strict mode
      #raise AssertionError('Out of range')
    return chr(i)

  elif id_ == Id.Char_Octal4:  # \0377 for echo -e
    # TODO: Error checking for \0777
    s = value[2:]
    i = int(s, 8)

gold/dollar-sq.sh

@@ -17,10 +17,15 @@ echo $'abcd\x65f'
echo $'abcd\044e'
echo $'abcd\u0065f'
echo $'abcd\U00000065f'
# In bash, these are different than echo -e.  I'm not sure why yet.
#echo $'\03777' | od -A n -t x1 | sed 's/ \+/ /g'
#echo $'\04000' | od -A n -t x1 | sed 's/ \+/ /g'
#echo $'\0777' | od -A n -t x1 | sed 's/ \+/ /g'

# NOTE: $'\377' is echo -e '\0377', with leading 0
echo $'\3777' | od -A n -t x1 | sed 's/ \+/ /g'
echo $'\4010' | od -A n -t x1 | sed 's/ \+/ /g'
echo $'\777' | od -A n -t x1 | sed 's/ \+/ /g'

# This wraps to \0, so it's not used.
#echo $'\4000' | od -A n -t x1 | sed 's/ \+/ /g'

echo $'abcd\x6' | od -A n -c | sed 's/ \+/ /g'
echo $'\x' $'\xg' | od -A n -c | sed 's/ \+/ /g'
echo $'abcd\04' | od -A n -c | sed 's/ \+/ /g'

osh/lex.py

@@ -395,9 +395,6 @@ def IsKeyword(name):

# Shared between echo -e and $''.
_C_STRING_COMMON = [
  # Note: tokens above \0377 can either be truncated or be flagged a syntax
  # error in strict mode.
  R(r'\\0[0-7]{1,3}', Id.Char_Octal),

  # \x6 is valid in bash
  R(r'\\x[0-9a-fA-F]{1,2}', Id.Char_Hex),
@@ -412,6 +409,10 @@ def IsKeyword(name):

# Used by ECHO_LEXER in core/builtin.py.
ECHO_E_DEF = _C_STRING_COMMON + [
  # Note: tokens above \0377 can either be truncated or be flagged a syntax
  # error in strict mode.
  R(r'\\0[0-7]{1,3}', Id.Char_Octal4),

  C(r'\c', Id.Char_Stop),

  # e.g. \A -> \A, is not a backslash escape.
@@ -427,15 +428,14 @@ def IsKeyword(name):
  R(r'[^\\]+', Id.Char_Literals),
]

UNUSED = [
  # For re2c.  TODO: need to make that translation.
  C('\\\0', Id.Char_Literals),
]

# NOTE: Id.Ignored_LineCont is also not supported here, even though the whole
# point of it is that supports other backslash escapes like \n!  It just
# becomes a regular backslash.
LEXER_DEF[lex_mode_e.DOLLAR_SQ] = _C_STRING_COMMON + [
  # Silly difference!  In echo -e, the syntax is \0377, but here it's $'\377',
  # with no leading 0.
  R(r'\\[0-7]{1,3}', Id.Char_Octal3),

  # \' is escaped in $'' mode, but not echo -e
  C(r'\'', Id.Char_OneChar),

spec/quote.test.sh

@@ -145,6 +145,19 @@ echo $'col1\ncol2\ncol3'
# In dash, \n is special within single quotes
# N-I dash stdout-json: "$col1\ncol2\ncol3\n"

### $'' octal escapes don't have leading 0
# echo -e syntax is echo -e \0377
echo -n $'\001' $'\377' | od -A n -c | sed 's/ \+/ /g'
## STDOUT:
 001 377
## END
## N-I dash STDOUT:
 $ 001 $ 377
## END
## BUG ash STDOUT:
 001 0O7
## END

### $""
echo $"foo"
# stdout: foo