Created: (HTTPCLIENT-588) relative URIs with internal double-slashes ('//') misparsed

relative URIs with internal double-slashes ('//')
misparsed
-----------------------------------------------------------

         Key: HTTPCLIENT-588
         URL: h
ttp://issues.apache.org/jira/browse/HTTPCLIENT-588
     Project: Jakarta HttpClient
        Type: Bug

  Components: HttpClient  
    Versions: 3.0.1    
    Reporter: Gordon Mohr


URI.parseUriReference()'s heuristic for interpreting URI
parts is thrown off by relative URIs which include an
internal '//'. As a result, portions of the supplied
relative URI (path) can be lost. 

For example:

URI rel = new URI("foo//bar//baz");
rel.toString();
(java.lang.String) //bar//baz

The culprit seems to be line 1961 of URI improperly
concluding that two slashes later than the beginning of
'tmp' are still indicative the URI is a 'net_path'. 

A possible quick fix might be to add a '!isStartedFromPath
&&' to the beginning of the line 1961 test, making
the line:

            if (!isStartedFromPath && at + 2 <
length && tmp.charAt(at + 1) == '/') {

... and thus preventing the misguided authority-parsing from
happening when earlier analysis already identified the
current string as a strictly path-oriented URI.

(It also appears the setting of the is_net_path boolean at
the end of this if's block may be wrong; this code is run
for hier_path URIs that are not net_paths in the 2396
syntax. For example:

URI uri = new URI("http://www.example.c
om/some/page");
uri.isNetPath();
 (boolean) true 

)

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the
administrators:
   http://issues.apache.org/jira/secure/Administrators.jspa

-
For more information on JIRA, see:
   http://www.atl
assian.com/software/jira


------------------------------------------------------------
---------
To unsubscribe, e-mail: httpclient-dev-unsubscribejakarta.apache.org
For additional commands, e-mail: httpclient-dev-helpjakarta.apache.org

    [ http://issues.apache.org/jira/b
rowse/HTTPCLIENT-588?page=comments#action_12416525 ] 

Oleg Kalnichevski commented on HTTPCLIENT-588:
----------------------------------------------

We will happily accept a patch

Oleg

> relative URIs with internal double-slashes ('//')
misparsed
>
-----------------------------------------------------------
>
>          Key: HTTPCLIENT-588
>          URL: h
ttp://issues.apache.org/jira/browse/HTTPCLIENT-588
>      Project: Jakarta HttpClient
>         Type: Bug

>   Components: HttpClient
>     Versions: 3.0.1
>     Reporter: Gordon Mohr

>
> URI.parseUriReference()'s heuristic for interpreting
URI parts is thrown off by relative URIs which include an
internal '//'. As a result, portions of the supplied
relative URI (path) can be lost. 
> For example:
> URI rel = new URI("foo//bar//baz");
> rel.toString();
> (java.lang.String) //bar//baz
> The culprit seems to be line 1961 of URI improperly
concluding that two slashes later than the beginning of
'tmp' are still indicative the URI is a 'net_path'. 
> A possible quick fix might be to add a
'!isStartedFromPath &&' to the beginning of the
line 1961 test, making the line:
>             if (!isStartedFromPath && at + 2
< length && tmp.charAt(at + 1) == '/') {
> ... and thus preventing the misguided authority-parsing
from happening when earlier analysis already identified the
current string as a strictly path-oriented URI.
> (It also appears the setting of the is_net_path boolean
at the end of this if's block may be wrong; this code is
run for hier_path URIs that are not net_paths in the 2396
syntax. For example:
> URI uri = new URI("http://www.example.c
om/some/page");
> uri.isNetPath();
>  (boolean) true 
> )

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the
administrators:
   http://issues.apache.org/jira/secure/Administrators.jspa

-
For more information on JIRA, see:
   http://www.atl
assian.com/software/jira


------------------------------------------------------------
---------
To unsubscribe, e-mail: httpclient-dev-unsubscribejakarta.apache.org
For additional commands, e-mail: httpclient-dev-helpjakarta.apache.org

     [ http://issues.apache.org/jira/browse/HTTPCLIENT-588?
page=all ]

Gordon Mohr updated HTTPCLIENT-588:
-----------------------------------

    Attachment: httpclient-588.patch

Attached patch avoids double-slash-triggered parsing of
authority when it would be inappropriate: when URI is a
simple authority-less path. Test for setting
'isStartedWithPath' improved to not indicate 'true' for
URIs that begin '//'; 'isStartedWithPath' reused in
later check to suppress inappropriate authority-parsing.
Unit test verifies problem in current code and corrected
behavior for test case after fix. 

> relative URIs with internal double-slashes ('//')
misparsed
>
-----------------------------------------------------------
>
>          Key: HTTPCLIENT-588
>          URL: h
ttp://issues.apache.org/jira/browse/HTTPCLIENT-588
>      Project: Jakarta HttpClient
>         Type: Bug

>   Components: HttpClient
>     Versions: 3.0.1
>     Reporter: Gordon Mohr
>  Attachments: httpclient-588.patch
>
> URI.parseUriReference()'s heuristic for interpreting
URI parts is thrown off by relative URIs which include an
internal '//'. As a result, portions of the supplied
relative URI (path) can be lost. 
> For example:
> URI rel = new URI("foo//bar//baz");
> rel.toString();
> (java.lang.String) //bar//baz
> The culprit seems to be line 1961 of URI improperly
concluding that two slashes later than the beginning of
'tmp' are still indicative the URI is a 'net_path'. 
> A possible quick fix might be to add a
'!isStartedFromPath &&' to the beginning of the
line 1961 test, making the line:
>             if (!isStartedFromPath && at + 2
< length && tmp.charAt(at + 1) == '/') {
> ... and thus preventing the misguided authority-parsing
from happening when earlier analysis already identified the
current string as a strictly path-oriented URI.
> (It also appears the setting of the is_net_path boolean
at the end of this if's block may be wrong; this code is
run for hier_path URIs that are not net_paths in the 2396
syntax. For example:
> URI uri = new URI("http://www.example.c
om/some/page");
> uri.isNetPath();
>  (boolean) true 
> )

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the
administrators:
   http://issues.apache.org/jira/secure/Administrators.jspa

-
For more information on JIRA, see:
   http://www.atl
assian.com/software/jira


------------------------------------------------------------
---------
To unsubscribe, e-mail: httpclient-dev-unsubscribejakarta.apache.org
For additional commands, e-mail: httpclient-dev-helpjakarta.apache.org

     [ http://issues.apache.org/jira/browse/HTTPCLIENT-588?
page=all ]

Oleg Kalnichevski updated HTTPCLIENT-588:
-----------------------------------------

    Fix Version: 3.1 Alpha 1

Seems all right. I'll check the patch in in a few days if
nobody complains

Oleg

> relative URIs with internal double-slashes ('//')
misparsed
>
-----------------------------------------------------------
>
>          Key: HTTPCLIENT-588
>          URL: h
ttp://issues.apache.org/jira/browse/HTTPCLIENT-588
>      Project: Jakarta HttpClient
>         Type: Bug

>   Components: HttpClient
>     Versions: 3.0.1
>     Reporter: Gordon Mohr
>      Fix For: 3.1 Alpha 1
>  Attachments: httpclient-588.patch
>
> URI.parseUriReference()'s heuristic for interpreting
URI parts is thrown off by relative URIs which include an
internal '//'. As a result, portions of the supplied
relative URI (path) can be lost. 
> For example:
> URI rel = new URI("foo//bar//baz");
> rel.toString();
> (java.lang.String) //bar//baz
> The culprit seems to be line 1961 of URI improperly
concluding that two slashes later than the beginning of
'tmp' are still indicative the URI is a 'net_path'. 
> A possible quick fix might be to add a
'!isStartedFromPath &&' to the beginning of the
line 1961 test, making the line:
>             if (!isStartedFromPath && at + 2
< length && tmp.charAt(at + 1) == '/') {
> ... and thus preventing the misguided authority-parsing
from happening when earlier analysis already identified the
current string as a strictly path-oriented URI.
> (It also appears the setting of the is_net_path boolean
at the end of this if's block may be wrong; this code is
run for hier_path URIs that are not net_paths in the 2396
syntax. For example:
> URI uri = new URI("http://www.example.c
om/some/page");
> uri.isNetPath();
>  (boolean) true 
> )

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the
administrators:
   http://issues.apache.org/jira/secure/Administrators.jspa

-
For more information on JIRA, see:
   http://www.atl
assian.com/software/jira


------------------------------------------------------------
---------
To unsubscribe, e-mail: httpclient-dev-unsubscribejakarta.apache.org
For additional commands, e-mail: httpclient-dev-helpjakarta.apache.org

    [ http://issues.apache.org/jira/b
rowse/HTTPCLIENT-588?page=comments#action_12416974 ] 

Roland Weber commented on HTTPCLIENT-588:
-----------------------------------------

looks good

cheers,
  Roland

> relative URIs with internal double-slashes ('//')
misparsed
>
-----------------------------------------------------------
>
>          Key: HTTPCLIENT-588
>          URL: h
ttp://issues.apache.org/jira/browse/HTTPCLIENT-588
>      Project: Jakarta HttpClient
>         Type: Bug

>   Components: HttpClient
>     Versions: 3.0.1
>     Reporter: Gordon Mohr
>      Fix For: 3.1 Alpha 1
>  Attachments: httpclient-588.patch
>
> URI.parseUriReference()'s heuristic for interpreting
URI parts is thrown off by relative URIs which include an
internal '//'. As a result, portions of the supplied
relative URI (path) can be lost. 
> For example:
> URI rel = new URI("foo//bar//baz");
> rel.toString();
> (java.lang.String) //bar//baz
> The culprit seems to be line 1961 of URI improperly
concluding that two slashes later than the beginning of
'tmp' are still indicative the URI is a 'net_path'. 
> A possible quick fix might be to add a
'!isStartedFromPath &&' to the beginning of the
line 1961 test, making the line:
>             if (!isStartedFromPath && at + 2
< length && tmp.charAt(at + 1) == '/') {
> ... and thus preventing the misguided authority-parsing
from happening when earlier analysis already identified the
current string as a strictly path-oriented URI.
> (It also appears the setting of the is_net_path boolean
at the end of this if's block may be wrong; this code is
run for hier_path URIs that are not net_paths in the 2396
syntax. For example:
> URI uri = new URI("http://www.example.c
om/some/page");
> uri.isNetPath();
>  (boolean) true 
> )

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the
administrators:
   http://issues.apache.org/jira/secure/Administrators.jspa

-
For more information on JIRA, see:
   http://www.atl
assian.com/software/jira


------------------------------------------------------------
---------
To unsubscribe, e-mail: httpclient-dev-unsubscribejakarta.apache.org
For additional commands, e-mail: httpclient-dev-helpjakarta.apache.org

     [ http://issues.apache.org/jira/browse/HTTPCLIENT-588?
page=all ]
     
Oleg Kalnichevski resolved HTTPCLIENT-588:
------------------------------------------

    Resolution: Fixed

Patch checked in. Many thanks, Gordon

Oleg

> relative URIs with internal double-slashes ('//')
misparsed
>
-----------------------------------------------------------
>
>          Key: HTTPCLIENT-588
>          URL: h
ttp://issues.apache.org/jira/browse/HTTPCLIENT-588
>      Project: Jakarta HttpClient
>         Type: Bug

>   Components: HttpClient
>     Versions: 3.0.1
>     Reporter: Gordon Mohr
>      Fix For: 3.1 Alpha 1
>  Attachments: httpclient-588.patch
>
> URI.parseUriReference()'s heuristic for interpreting
URI parts is thrown off by relative URIs which include an
internal '//'. As a result, portions of the supplied
relative URI (path) can be lost. 
> For example:
> URI rel = new URI("foo//bar//baz");
> rel.toString();
> (java.lang.String) //bar//baz
> The culprit seems to be line 1961 of URI improperly
concluding that two slashes later than the beginning of
'tmp' are still indicative the URI is a 'net_path'. 
> A possible quick fix might be to add a
'!isStartedFromPath &&' to the beginning of the
line 1961 test, making the line:
>             if (!isStartedFromPath && at + 2
< length && tmp.charAt(at + 1) == '/') {
> ... and thus preventing the misguided authority-parsing
from happening when earlier analysis already identified the
current string as a strictly path-oriented URI.
> (It also appears the setting of the is_net_path boolean
at the end of this if's block may be wrong; this code is
run for hier_path URIs that are not net_paths in the 2396
syntax. For example:
> URI uri = new URI("http://www.example.c
om/some/page");
> uri.isNetPath();
>  (boolean) true 
> )

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the
administrators:
   http://issues.apache.org/jira/secure/Administrators.jspa

-
For more information on JIRA, see:
   http://www.atl
assian.com/software/jira


------------------------------------------------------------
---------
To unsubscribe, e-mail: httpclient-dev-unsubscribejakarta.apache.org
For additional commands, e-mail: httpclient-dev-helpjakarta.apache.org