Re: [patch #4598] filesystemRepresentation on Win32 is DANGEROUS as implemented

[Sheldon Gill]

Jeremy Bettis wrote:
> URL:
>   <http://savannah.gnu.org/patch/?func=detailitem&item_id=4598>
>
>                  Summary: filesystemRepresentation on Win32 is DANGEROUS as
> implemented
>                  Project: GNUstep
>             Submitted by: jeremybettis
>             Submitted on: Fri 11/04/05 at 21:01
>                 Category: Foundation
>                 Priority: 9 - Immediate
>                   Status: None
>                  Privacy: Public
>              Assigned to: None
>         Originator Email: 
>              Open/Closed: Open
>
>     _______________________________________________________
>
> Details:
>
> *) filesystemRepresentation is declared as returning char* when it actually
> returns unichar*
> *) Even gnustep-base itself has uses of filesystemRepresentation that are not
> expecting unichars.
> *) Unicode is NOT the native filesystem encoding for the Win32 API.  Win32
> actually has 2 "native" encodings.
>
> This patch resolves all of these issues.
> new method -unicodeFilesystemRepresentation
> fixed all win32 uses of filesystemRepresentation to use the new
> unicodeFilesystemRepresentation
> fixed all of the occurances of a unicode string being passed to a ANSI api
> call

I agree that fileSystemRepresentation as it currently is in HEAD is dangerous. 
However, I think this is the wrong solution.

fileSystemRepresentation should return char* so it can be passed on to *nix/C 
routines. That is what it was designed for.

We don't need to use it for unicode. It's symptomatic of the the current 
situation where Win32 unicode work is incomplete. We should fix the problems 
where they really lie rather than try to mask them with an additional method.

The difficulty here has arisen because of the path munging on Win32 which is 
really unnecessary and causes more problems than it's worth. We shouldn't need 
fileSystemRepresentation to convert from an internal format.



Regards,
Sheldon