[LLVMdev] getDirectoryContents and renameFile needs to be implemented in Win32/Path.cpp

Jeff Cohen jeffc at jolt-lang.org
Thu Dec 30 22:01:33 PST 2004

Reid Spencer wrote:

>On Thu, 2004-12-30 at 20:46, Jeff Cohen wrote:
>>And now that I've done that I can't help but notice that Path.cpp is 
>>getting very Unix-centric in its interfaces, what with file permission 
>>bits and stuff.  This is going to cause problems if and when LLVM is 
>>ported to other operating systems, though I have to admit it's getting 
>>difficult to name viable operating systems it could get ported to that 
>>isn't some flavor of Unix.  But there are a few left...
>Yeah, this is a known problem. I've been purposely holding off until all
>the Path functionality was done. That's complete now so I'm starting to
>think about making the interface less sucky. We're probably going to
>model it after java.io.File which has a pretty clean interface. As for
>the permission bits, this is just needed for llvm-ar whose file format
>dictates that it have mode bits. There might be some opportunity for
>cleaning that up as well. If you have any specific suggestions for
>improving the Path interface, I'm all ears.
>Also, thanks for completing the Win32 Path implementation.
Sigh... maybe it would be best to declare there shall not be a port to a 
non-Unix OS, except for Windows.  There really are few alternatives 
left.  The only ones I can think of that still have life are IBM's, and 
we don't want to go there.  An MVS port for IBM's mainframes?  I can't 
even begin to list the assumptions in Path.cpp that are false on that 
platform, and, unfortunately, it's not from ignorance. If it can't 
emulate the POSIX api, forget it--and I bet there's POSIX emulation for 
MVS nowadays anyway. 

More information about the llvm-dev mailing list