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

[Sheldon Gill]

Jeremy Bettis wrote:
> Follow-up Comment #1, patch #4598 (project gnustep):
>
> Ok, so how about replacing every occurance of unicodeFileSystemRepresentation
> with (unichar*)[path cStringUsingEncoding: NSUnicodeStringEncoding]
> which is what unicodeFileSystemRepresentation actually does internally.

Although fine for the moment, I don't think it's the best solution ultimately. 
We can clean up the code so unicode on Win32 is the norm and used everywhere.

I'd much rather see a transition to using Win32 API calls for this handling 
rather than routing it through the *nix/C compatibility layer. We're a little 
bit of the way there but there is plenty still to do.

> I would be fine with that approach also.  My non-negotiable is that
> filesystemRep returns char*.

Absolutely. The idea has always been:
    ret = nix_system_call( [pathString fileSystemRepresentation] );

and breakage of this shouldn't be accepted.

I think the goal, though, should be:

- for application code to rely entirely on NSString to hold the path and other 
NS API calls to operate on those files. Avoid the need to call down

- for core code to better encapsulate Win32 and use less *nix approaches. We 
gain in both performance and behaviour.


Regards,
Sheldon