Re: [Mingw-users] Where is __dirstream_t really defined?

[Keith M. <kei...@us...>]

On 06/02/15 18:08, Miro Kropáček wrote:
> Well, it's not my project, it's someone else's. It's used like this:
> 
> #define DIRHANDLE dirp->dd_handle
> 
>   DIR *dirp;
> 
>   if ((dirp = opendir(dirname)) == NULL)
>     return(-1);
> 
>   if (fstat(DIRHANDLE, &stb) < 0)
>     return(-1);
> 
> Not saying it's a great practice, though. So I'd like to make it compilable
> somehow, it works for other Unix flavors, incl. Cygwin.

Really?  You amaze me.  Can you name any such Unix flavours?  (FWIW,
GNU/Linux certainly isn't among them).

I see two fatally flawed assumptions in the above code snippet:

1) It assumes that the opaque entity referenced by dirp represents a
   struct or a union, which includes a dd_handle member; there is no
   way in which you can be certain of that.

2) Even if (1) were true, it assumes that this dd_handle member is a
   descriptor reference to a file stream; on Windows, it is impossible
   to process a directory as a regular file stream, (notwithstanding
   the hoops cygwin may jump through, to make it appear otherwise).

Like Eli says, your code snippet is not portable; in fact, it is just
plain broken.

-- 
Regards,
Keith.