Re: [Boost-users] [Spirit.Qi] failing to parse IP address octets

27 Nov 2017

      Just taking a 20k meter view of things, this is looking to me like a
bug in one or more of the following, or some Skipper dynamic I'm
unaware of:

* optional (-)
* alternative (|)

A bug in one or more of these, or others, piling up on each other.

Because, to me at least, the grammar looks fine, as far as I am aware,
with the Spirit Qi operators I know about.

Can anyone at all look into this?

Thank you...

On Mon, Nov 27, 2017 at 5:48 PM, Michael Powell  wrote:
...
On Mon, Nov 27, 2017 at 5:04 PM, Michael Powell  wrote:
...
On Mon, Nov 27, 2017 at 4:39 PM, Michael Powell  wrote:
...
Hello,
I've got the following parsers which are failing to find the correct attributes.
I tried to parse in terms of uint_parser() at
first, but this was not working properly.
So I tried the following:
_octet_rule = lexeme[
    char_('0')
        | (char_('1') >> -(digit >> -digit))
        | (char_('2')
            >> -((char_('0', '4') >> -(digit))
                | (char_('5') >> -(char_('0', '5')))
                | char_('6', '9')
                )
            )
        | (char_('3', '9') >> -digit)
];
And if I "flatten" the alternatives, I end up with a "match", but the
attributed match looks really odd, which is a post-parser error of its
own:
lexeme[
    char_('0')
        | (char_('1') >> -(digit >> -digit))
        | (char_('3', '9') >> -digit)
        | (char_('2') >> char_('5') >> -char_('0', '5'))
        | (char_('2') >> char_('0', '4') >> -digit)
        | (char_('2') >> char_('6', '9'))
        //| (char_('2')
        //    >> -(
        //    (char_('0', '4') >> -digit)
        //        | (char_('5') > -char_('0', '5'))
        //        | char_('6', '9')
        //        )
        //    )
];
The Address rule itself has not changed, is still incorporating this
Octet rule, with test failures:
FAILED:
  CHECK_THAT( y, Equals(x) )
with expansion:
  "84.2244.31.76" equals: "84.244.31.76"
       ^^^^^^^
with message:
  Verifying address >>> 84.244.31.76 <<<
FAILED:
  CHECK_THAT( y, Equals(x) )
with expansion:
  "251.170.2248.119" equals: "251.170.248.119"
               ^^^^^^^
with message:
  Verifying address >>> 251.170.248.119 <<<
FAILED:
  CHECK_THAT( y, Equals(x) )
with expansion:
  "87.119.108.224" equals: "87.119.108.24"
                   ^^^^^^^
with message:
  Verifying address >>> 87.119.108.24 <<<
FAILED:
  CHECK_THAT( y, Equals(x) )
with expansion:
  "160.3.2230.143" equals: "160.3.230.143"
            ^^^^^^^
with message:
  Verifying address >>> 160.3.230.143 <<<
What am I missing here? I'm sure I probably need to invoke something
like hold[] or lexeme[], beyond what the Octet pattern itself is
already doing?
...
...
Which is presently failing to find all three digits in the
"20\d"-"24\d" use case.
Octet itself parses correctly when I use this rule:
_octet_rule >> !(char_);
My octet test cases verify the entire range of 0-255 successfully, as
well as the following invalid cases:
VerifyOctet(-1, invalid_);
VerifyOctet(1990, invalid_);
VerifyOctet(2550, invalid_);
VerifyOctet("255.", invalid_);
VerifyOctet(256, invalid_);
VerifyOctet(999, invalid_);
VerifyOctet("abc", invalid_);
VerifyOctet(123.456, invalid_);
Converted to corresponding string, i.e. "1990" or "123.456".
But when I combine that into the IP address rule, the test results occur:
_octet_rule >> repeat(3)[char_('.') >> _octet_rule];
And the test results:
CHECK_THAT( y, Equals(x) )
with expansion:
  "145.23.183.47" equals: "145.231.183.47"
with message:
  Verifying address >>> 145.231.183.47 <<<
FAILED:
  CHECK_THAT( y, Equals(x) )
with expansion:
  "189.21.29.11" equals: "189.211.29.11"
with message:
  Verifying address >>> 189.211.29.11 <<<
FAILED:
  CHECK_THAT( y, Equals(x) )
with expansion:
  "90.60.32.24" equals: "90.60.32.246"
with message:
  Verifying address >>> 90.60.32.246 <<<
FAILED:
  CHECK_THAT( y, Equals(x) )
with expansion:
  "22.59.144.140" equals: "223.59.144.140"
with message:
  Verifying address >>> 223.59.144.140 <<<
Perhaps I should have more test cases represented as well, because
when the 2's is the single digit, I get a space interpreted. Which is
starting to look like a "skipper" issue to me?
FAILED:
  CHECK_THAT( y, Equals(x) )
with expansion:
  "22.23.2 .152" equals: "221.236.2.152"
            ^^^^
with message:
  Verifying address >>> 221.236.2.152 <<<
...
IP address test data is randomly generated, or combinatorially if I
wanted to wait for all 4.3B or so test cases. However, I figure that
as long as the Octet rule is working, it is reasonable to expect the
Address rule may also work.
Not sure why those last digits are being dropped, however. Or how the
1's can succeed where the 2's are failing.
Any suggestions, Spirit folks?
Thanks!
Cheers,
Michael Powell