Bug of cvtenc (Re: ANN: GNUstep base/make Version 1.7.4)

[Yen-Ju Chen]

> From: Adam Fedor <fedor@doc.com>
> To: Yen-Ju Chen <yjchenx@hotmail.com>
> CC: "bug-gnustep@gnu.org" <bug-gnustep@gnu.org>
> Subject: Re: ANN: GNUstep base/make Version 1.7.4
> Date: 20 Sep 2003 09:27:07 -0600
>
> On Fri, 2003-09-19 at 22:20, Yen-Ju Chen wrote:
>
> I don't understand. If eIn=YES, then it does use iEnc. I didn't change
> anything about how the code read the input. If the input is in Unicode,
> then the code correctly determines that, otherwise it assumes it is in
> the default encoding (even though it may just be escaped unicode). In
> that case you need to specify EscapeIn or EscapeOut to tell it which is
> which.

 O.K. For "-EscapeIn yes" (eIn == YES) case.
 First it tests whether it is Unicode (0xFFFE || oxFEFF),
 if so, iEnc = Unicode, oEnc is local encoding.
 if not, iEnc = local encoding, oEnc is Unicode.
 Then convert NSData to NSString using iEnc, which is correct.

 Later, it convert evey \uXXXX into unichar in NSString.

 Finally, it write the NSString to NSData using oEnc, then write to file.

 The last step is wrong because it always use oEnc for writting.
 In the case of iEnc = Unicode, it will be writted as local encoding.
 In the case of iEnc = local encoding, it will be written as Unicode.
 It is exactly opposite what user want.
 User only want to use cvtenc to convert \uXXXX to their local character
 so that they can easily modify the property list in local encoding 
environment.
 In the case, cvtenc should read and write in local encoding,
 only convert \uXXXX to real character.

 It's not about Unicode or not.
 It's about cvtenc should use the same encoding for read and write.
 \uXXXX is not Unicode. It is the presentation of Unicode in local 
encoding.

> Do you have a test file that will show me what is wrong?

 I can send you a test in traditional Chinese (Big5 Encoding),
 but it probably looks like unicode for you. :)

 You can simply take any language in 
SYSTEM/Library/Libraries/Resource/gnustep-base/Languages.
 And according to the README in it,
 you can use "cvtenc -EscapeIn yes French > tmpfile" to get a local 
encoding file (ISO8895-1 ?).
 From my test, you will get a Unicode (UCS-Internal or other UCS) file.

 Yen-Ju

> _______________________________________________
> Bug-gnustep mailing list
> Bug-gnustep@gnu.org
> http://mail.gnu.org/mailman/listinfo/bug-gnustep

_________________________________________________________________
Instant message with integrated webcam using MSN Messenger 6.0. Try it now 
FREE!  http://msnmessenger-download.com