190749 - IRC's CASEMAPPING is not ASCII or latin-1 (except for dalnet)

Status: RESOLVED FIXED

(Other Applications :: ChatZilla, defect)

Description

[Perry Lorier]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2) Gecko/20021126
Build Identifier: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2) Gecko/20021126

When IRC was first invented it was used in finland, so used a finnish
characterset.  This means it's not your usual ASCII.  In addition to the usual
A-Z being equivilent to a-z, ~ is equivilent to ^, [ and {, ] and }, | and \.

This is for channels and nicks.  Note that "~" isn't valid in a nick, but "^" is.

This of course causes issues if you joined #\foo\, but messages come to the
client that are destined to #|foo|

Reproducible: Always

Steps to Reproduce:
1. take two clients, join one to #\foo\
2. join the other to #|foo|
3. talk in the channel from both and hilarity ensues.



Expected Results:  
mozilla should treat equivilent characters as equivilent.

http://www.alien.net.au/irc/index.html has information on 005 which explains
about the CASEMAPPING 005 parameter for clients so they can tell which
transformations the server is using.

RFC1459 also documents some of these mappings, but misses ~ being equivilent to ^.

Comment 1

[Perry Lorier]

the 005 draft is:
http://www.ietf.org/internet-drafts/draft-brocklesby-irc-isupport-01.txt

Comment 2

[Samuel Sieb]

ugh :-/

Comment 3

[James Ross]

I'm working on a general 005 patch, so we'll have all the information available
to whatever needs it - I think this bug could be done using some
toLowerCase/toUpperCase functions set up to use the correct mappings. Might try
this in a few days, depending how the 005 patch goes.

Comment 4

[James Ross]

Seems the 005 draft has been updated, it's now here:
http://www.ietf.org/internet-drafts/draft-brocklesby-irc-isupport-02.txt

Comment 5

[James Ross]

The RPL_isupport code has now landed, so this should be relatively easy, since
all of the case-conversation is very well defined.

Comment 6

[James Ross]

Working on this, got it mostly working, though I'm being hindered a bit by one
of the two servers I use completely failing to follow any sort of spec (moznet
is fine, of course ;) ).

Comment 7

[James Ross]

Attachment: Use casemapping-correct toLowerCase conversions.

This patch defines a toLowerCase function on the CIRCServer object, which obays
the current casemapping in effect. It also makes the appropriate calls to this
function from the various objects. Tab-complection has also been tweaked to
work with it, which is very nice to use.

Comment 8

[James Ross]

Comment on attachment 141184 [details] [diff] [review]
Use casemapping-correct toLowerCase conversions.

Probably should be looked over by Robert Ginda too, if time allows.

Comment 9

[Samuel Sieb]

Comment on attachment 141184 [details] [diff] [review]
Use casemapping-correct toLowerCase conversions.

Your switch statement doesn't have any breaks;

'if (d.server)' should be 'if "server" in d'

Comment 10

[Samuel Sieb]

Comment on attachment 141184 [details] [diff] [review]
Use casemapping-correct toLowerCase conversions.

My mistake.
Fix the cases and r=me

Comment 11

[James Ross]

Comment on attachment 141184 [details] [diff] [review]
Use casemapping-correct toLowerCase conversions.

Checked in.

Comment 12

[James Ross]

I believe the patch just checked in covers this bug, so resolving FIXED. If this
is not the case, please reopen.