57104 - Text incorrectly interpreted as a url. (Regex error)

Status: VERIFIED FIXED

(Other Applications Graveyard :: ChatZilla, defect, P3)

Description

[David Krause]

Overview Description:
Text with the characters "www..." is incorrectly interpreted as a url.

Steps to Reproduce:
1. Start ChatZilla and join a channel.
2. Type "www..."
3. Observe

Actual Results:
The text is converted to a link.

Expected Results:
No link.

Reproducibility:
everytime

Build Date & Platform Bug Found:
Mozilla/5.0 (X11; U; Linux 2.2.16-3 i686; en-US; m18) Gecko/20001012

Additional Builds and Platforms Tested On:
none

Additional Information:

I believe the relevant code is
http://lxr.mozilla.org/seamonkey/source/extensions/irc/xul/content/static.js#204

obj.munger.addRule
("link",
/((\w+)\:\/\/[^\<\>\(\)\'\"\s]*|www\.[^\<\>\(\)\'\"\s]+\.[^\<\>\(\)\'\"\s]*)/,
insertLink);

Comment 1

[David Krause]

CC Hixie to see if he wants to have some regex fun. :)

Comment 2

[David Krause]

Found this nice little regex on the NLUG list archives.  Maybe it would be of
some use.

#!/usr/bin/perl

while (<>) {
if
(($service,$user,$password,$host,$port,$file,$params,$anchor)=(/^(?:([a-zA-A
]+)\:\/\/)?(?:([A-Za-z0-9\_\.\-]*?)(?:\:([A-Za-z0-9\_\.\-]*?))?\@)?([a-zA-Z0
-9\.\_\-]*?)(?:\:([0-9]{1,5}))?(\/.*?)?(?:\?(.*?))?(?:#(.*?))?$/)) {
print "Service: ${service}\nUser: ${user}\nPassword: ${password}\n";
print "Host: ${host}\nPort: ${port}\nFile: ${file}\n";
print "Params: ${params}\nAnchor: ${anchor}\n\n";
} else {
print "I don't recognize this as a URL.\n";
}
}

Comment 3

[timeless]

I don't think so.
D:\CVS\104\mozilla\themes\classic\communicator\mac>echo www.cnn.com | perl -e 
"while (<>) {if 
(($service,$user,$password,$host,$port,$file,$params,$anchor)=(/^(?:([a-zA-A]+)
\:\/\/)?(?:([A-Za-z0-9\_\.\-]*?)(?:\:([A-Za-z0-9\_\.\-]*?))?\@)?([a-zA-Z0-9\.\_
\-]*?)(?:\:([0-9]{1,5}))?(\/.*?)?(?:\?(.*?))?(?:#(.*?))?$/)) {print qq(Service: 
${service}\nUser: ${user}\nPassword: ${password}\n);print qq(Host: 
${host}\nPort: ${port}\nFile: ${file}\n);print qq(Params: ${params}\nAnchor: 
${anchor}\n\n);} else {print qq(I don't recognize this as a URL.\n);}}"
I don't recognize this as a URL.

according to a survey of people i respect .. nullifises domain names

my quick fix:
obj.munger.addRule
("link",
/((\w+)\:\/\/[^\<\>\(\)\'\"\s]*|www\.[^.\<\>\(\)\'\"\s]+\.[^.\<\>\(\)\'\"\s]*)/
,
insertLink);

Comment 4

[Decklin Foster]

Can we fix this while we're at it? we match "www.foo.com" but not
"www.foo.co.uk". For that matter, we match "www.foo."... Also, you do not need
to backwack metachars in character classes.

/((\w+):\/\/[^<>()'"\s]+|www(\.[^.<>()'"\s]+){2,})/

but preferably:

/((\w+):\/\/[^<>()'"\s]+|www(\.[-\w]+){2,})/

If the parens and choice of // as delimiters are not necessary I'd drop them for
readability's sake.

Comment 5

[timeless]

i'm thinking about the following:

www|ftp
[.{text}]*
.\w\w\w?

ie, ftp.mozilla.co, ftp.mozilla.com, not ftp.mozilla.come
this catches all countries, and current tlds. of course i'll need to fit this 
into the current scheme.

Comment 6

[Robert Ginda]

fix checked in

Comment 7

[David Krause]

Verified with 2000120808 linux build.