Opened 13 years ago

Closed 8 years ago

#182 closed defect (fixed)

iconv() and path separator

Reported by: guest Owned by: bird
Priority: normal Milestone: libc-0.6.5
Component: libc Version: 0.6
Severity: normal Keywords: iconv
Cc: komh@…



Current iconv() implementation does not recognize path separator. At least in Korea, '\' is used as currency, WON sign, as well. So if iconv() does not treat it as path separator, path icluding '\' is not recognized correctly. This is the cause of SAMBA client not working with path including DBCS chars and '\'.

So I think it is better that you let us make a choice for using '\' as path separator or others.

Of course, this proposal goes for CVTTYPE_CTRL7F.

That is, at first, you'd not better set attribute of uconv object, and later if requested, set attribute. Or you can use init string as UconvCreateObject?() does.

Anyway give us a choice right.

KO Myung-Hun

Attachments (1)

iconv_path.diff (1.1 KB) - added by KO Myung-Hun 9 years ago.
A patch to clear CVTTYPE_CTRL7F flag only

Download all attachments as: .zip

Change History (7)

comment:1 Changed 10 years ago by bird

I don't quite get this defect. In which way does iconv() need to care about path separators? A testcase explaining what's wrong and how it really should would be probably clarify things quite a bit (though I'm not sure if I can find any Korean OS/2 to test it on, if that's necessary). As always, even better if a patch with a fix is also attached.

comment:2 Changed 10 years ago by KO Myung-Hun

Cc: komh@… added

comment:3 Changed 10 years ago by KO Myung-Hun

For examples, if trying to convert A\B\C to unicode, the result is [41][5C][42][5C][43] in unicode with path mode, but [41][20a9][42][20a9][43] with non-path mode. [20a9] is a WON sign in unicode.

So, if you create a directory structure using the unicode string, you can get the directory whose name is A[WON]B[WON]C not a A - B - C directory tree.

And GNU iconv() treats both \ and 0x7F as an ASCII. That is, it converts them to [5C] and [7F] in unicode, respectively.

So clearing CVTTYPE_CTRL7F flag only is the way compatible with GNU iconv().

Changed 9 years ago by KO Myung-Hun

Attachment: iconv_path.diff added

A patch to clear CVTTYPE_CTRL7F flag only

comment:4 Changed 8 years ago by bird

Keywords: iconv added
Milestone: libc-0.6.5
Resolution: fixed
Status: newclosed
Version: 0.6

Changed iconv to not clear CVTTYPE_PATH in r3754. Backported to 0.6 in r3755. Case closed.

comment:5 Changed 8 years ago by KO Myung-Hun

Resolution: fixed
Status: closedreopened

Comment is wrong.

WON is assigned to an ASCII code of a backward slash('\') not a forward slash('/').

comment:6 Changed 8 years ago by bird

Resolution: fixed
Status: reopenedclosed

Fixed comment.

Note: See TracTickets for help on using tickets.