GNU bug report logs - #10410
guile: uri module confused by domain names starting with numbers, ipv6 addresses

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 10410 in the body.
You can then email your comments to 10410 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Daniel Hartwig <mandyke <at> gmail.com>
To: submit <at> debbugs.gnu.org
Subject: guile: uri module confused by domain names starting with numbers,
	ipv6 addresses
Date: Fri, 30 Dec 2011 18:14:56 +0800

[Message part 1 (text/plain, inline)]

Package: guile
Version: 2.0.3
Tags: patch
X-Debbugs-CC: guile-devel <at> gnu.org


Hello

I have noticed that the (web uri) module does not handle domain names
that start with numbers:

scheme@(guile-user)> (string->uri "http://123.com")
$1 = #f
scheme@(guile-user)> (build-uri 'http #:host "123.com")
web/uri.scm:85:6: In procedure build-uri:
web/uri.scm:85:6: Throw to key `uri-error' with args `("Expected valid
host: ~s" ("123.com"))'.


Also, `string->uri' does not handle ipv6 addresses:

scheme@(guile-user)> (string->uri "http://[2001:db8::1]")
$2 = #f


Attached patch implements support for domain names that start with
numbers by correcting the
regular expressions used by `valid-host?' as well as some related tests.

`string->uri' requires similar changes to support the ipv6 address
literals.  I'm yet to found a very elegant way to do this though it is
easy enough to simply butcher `authority-pat'.

[0001-support-URIs-with-domain-names-starting-with-numbers.patch (text/x-patch, attachment)]

Message #8 received at 10410 <at> debbugs.gnu.org (full text, mbox):

From: Daniel Hartwig <mandyke <at> gmail.com>
To: 10410 <at> debbugs.gnu.org
Cc: guile-devel <at> gnu.org
Subject: Re: bug#10410: guile: uri module confused by domain names starting
	with numbers, ipv6 addresses
Date: Sat, 31 Dec 2011 00:27:53 +0800

[Message part 1 (text/plain, inline)]

On 30 December 2011 18:14, Daniel Hartwig <mandyke <at> gmail.com> wrote:
>
> `string->uri' requires similar changes to support the ipv6 address
> literals.  I'm yet to found a very elegant way to do this though it is
> easy enough to simply butcher `authority-pat'.

So the issue was really with `parse-authority'.

The attached patch cleans this up with support for IPv6 (including
dotted-quad notation), fixes some typos in the tests, and adds new
tests.

With both patches applied the web-uri.test now passes for all tests
and I can finally do:

scheme@(guile-user)> (string->uri "http://[::ffff:192.0.2.1]/foo")
$2 = #<<uri> scheme: http userinfo: #f host: "[::ffff:192.0.2.1]"
port: #f path: "/foo" query: #f fragment: #f>
scheme@(guile-user)> (string->uri "http://123.com")
$3 = #<<uri> scheme: http userinfo: #f host: "123.com" port: #f path:
"" query: #f fragment: #f>

[0002-enhance-IPv6-support.patch (text/x-patch, attachment)]

Message #11 received at 10410 <at> debbugs.gnu.org (full text, mbox):

From: Daniel Hartwig <mandyke <at> gmail.com>
To: 10410 <at> debbugs.gnu.org
Subject: bump: Bug#10410: numbers, ipv6 addresses in uri module
Date: Wed, 22 Feb 2012 14:15:26 +0800

Bumping this bug report as it is relevant to commit 1868309 which
recently addressed the issue with domain names starting with numbers.

> author        Andy Wingo <wingo <at> pobox.com>    2012-02-21 16:42:53 (GMT)
> commit        1868309a9e34a04a5b3020e147d0ce029038b290 (patch) (side-by-side diff)
> ...
> (define ipv6-regexp
> - (make-regexp "^\\[([0-9a-fA-F:]+)\\]+"))
> + (make-regexp "^\\[([0-9a-fA-F:]+)\\]+$"))

Please compare this regexp to the one in the previously sent 0001
patch.  You will note that the second "+" here matches values with
extraneous characters such as "[::1]]]".

Also, the character group does not include a "." which is valid in
ipv6 addresses with dotted-quad notation[1].  This is introduced by
the 0002 patch (see below).

0001 provides additional test cases handling domains names starting
with numbers as well.

Daniel Hartwig <mandyke <at> gmail.com> wrote:
> Also, `string->uri' does not handle ipv6 addresses:
>
> scheme@(guile-user)> (string->uri "http://[2001:db8::1]")
> $2 = #f

This issue remains outstanding and addressed in the 0002 patch with
test cases provided.

Regards


[1] http://tools.ietf.org/html/rfc1884#page-5

    3. An alternative form that is sometimes more convenient when
       dealing with a mixed environment of IPv4 and IPv6 nodes is
       x:x:x:x:x:x:d.d.d.d, where the 'x's are the hexadecimal values
       of the six high-order 16-bit pieces of the address, and the 'd's
       are the decimal values of the four low-order 8-bit pieces of the
       address (standard IPv4 representation).  Examples:

            0:0:0:0:0:0:13.1.68.3

            0:0:0:0:0:FFFF:129.144.52.38

       or in compressed form:

            ::13.1.68.3

            ::FFFF:129.144.52.38

Message #14 received at 10410 <at> debbugs.gnu.org (full text, mbox):

From: ludo <at> gnu.org (Ludovic Courtès)
To: Daniel Hartwig <mandyke <at> gmail.com>
Cc: 10410 <at> debbugs.gnu.org, guile-devel <at> gnu.org
Subject: Re: bug#10410: guile: uri module confused by domain names starting
	with numbers, ipv6 addresses
Date: Wed, 20 Jun 2012 15:39:30 +0200

Hi Daniel,

Daniel Hartwig <mandyke <at> gmail.com> skribis:

> I have noticed that the (web uri) module does not handle domain names
> that start with numbers:
>
> scheme@(guile-user)> (string->uri "http://123.com")
> $1 = #f

This one was fixed around commit 1868309a9e34a04a5b3020e147d0ce029038b290.

Thanks,
Ludo’.

Message #19 received at 10410-done <at> debbugs.gnu.org (full text, mbox):

From: Andy Wingo <wingo <at> pobox.com>
To: Daniel Hartwig <mandyke <at> gmail.com>
Cc: 10410-done <at> debbugs.gnu.org
Subject: Re: bug#10410: bump: Bug#10410: numbers, ipv6 addresses in uri module
Date: Fri, 06 Jul 2012 13:15:12 +0200

On Wed 22 Feb 2012 07:15, Daniel Hartwig <mandyke <at> gmail.com> writes:

> Bumping this bug report as it is relevant to commit 1868309 which
> recently addressed the issue with domain names starting with numbers.

Sorry for the long delay.  I have applied your patches on my hasty
commits.  There were a couple bugs that this brought out; for example on
my system, inet-pton does not expect the address to be surrounded by
[].  I fixed this to have the uri-host / build-uri #:host to never have
brackets.

WDYT?  Marking as done as your code is in, with tests.  Thanks very much
for the report and patch!

Andy
-- 
http://wingolog.org/

Message #22 received at 10410 <at> debbugs.gnu.org (full text, mbox):

From: Daniel Hartwig <mandyke <at> gmail.com>
To: 10410 <at> debbugs.gnu.org
Subject: Re: bug#10410: bump: Bug#10410: numbers, ipv6 addresses in uri module
Date: Sat, 7 Jul 2012 12:01:06 +0800

On 6 July 2012 19:15, Andy Wingo <wingo <at> pobox.com> wrote:
> On Wed 22 Feb 2012 07:15, Daniel Hartwig <mandyke <at> gmail.com> writes:
>
>> Bumping this bug report as it is relevant to commit 1868309 which
>> recently addressed the issue with domain names starting with numbers.
>
> Sorry for the long delay.

Please, no :-)  Your meticulous approach to maintenance is
commendable.  All are confident of the eventuality of the next “serial
processing” session.

> I have applied your patches on my hasty
> commits.  There were a couple bugs that this brought out; for example on
> my system, inet-pton does not expect the address to be surrounded by
> [].

Nice spot.  Evidently when I originally looked in to this I did not
get much past string->uri and assumed the job done (*what* happened to
the actual task I was trying to complete then?!)

> I fixed this to have the uri-host / build-uri #:host to never have
> brackets.

Works well.  One small nitpick, depending on what uri-host is defined
to be returning this may or may not differ from spec.:

 host        = IP-literal / IPv4address / reg-name

 IP-literal    = "[" ( IPv6address / IPvFuture  ) "]"

In my intuition, a call to build-uri is naturally of the form:

 (build-uri #:scheme 'http
            #:host "[2001:db8::1]")

Though, for convenience, this begs for a predicate or two for
ip-literals and conversion from ip-literal->ipv6-address.

Anyway, working is better than not working, thanks :-)

Message #25 received at 10410 <at> debbugs.gnu.org (full text, mbox):

From: Andy Wingo <wingo <at> pobox.com>
To: Daniel Hartwig <mandyke <at> gmail.com>
Cc: 10410 <at> debbugs.gnu.org
Subject: Re: bug#10410: bump: Bug#10410: numbers, ipv6 addresses in uri module
Date: Sat, 07 Jul 2012 11:38:02 +0200

Hi Daniel,

Thanks for testing (& for the original patches :)

On Sat 07 Jul 2012 06:01, Daniel Hartwig <mandyke <at> gmail.com> writes:

> One small nitpick, depending on what uri-host is defined to be
> returning this may or may not differ from spec.:
>
>  host        = IP-literal / IPv4address / reg-name
>
>  IP-literal    = "[" ( IPv6address / IPvFuture  ) "]"
>
> In my intuition, a call to build-uri is naturally of the form:
>
>  (build-uri #:scheme 'http
>             #:host "[2001:db8::1]")

Yeah it is a bit confusing.  I think the thing to do is to consider the
brackets as part of how ip-literals are delimited within a URI, and not
part of the host itself.  Thus it would be Guile's responsibility to add
them or remove them.  But as you say, this goes against the RFC.

At the same time, I'd like to be able to always pass the result of
inet-ntop to build-uri #:host.

The internet is a messy place!

Andy
-- 
http://wingolog.org/

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.