[Mono-list] Mono.Unix.Native.Syscall.(l)stat weirdness

[László Monda]

Hey Jonathan,

On Wed, 2006-02-15 at 07:15 -0500, Jonathan Pryor wrote:
> On Wed, 2006-02-15 at 12:15 +0100, László Monda wrote:
> > I've been playing with stat() and lstat() in the past few days and
> > things seem to be weird.
> > 
> > I made two test programs to demonstrate the problem.  The first is
> > written in C and behaves correctly, the second is written in C# and is
> > quite abnormal.  I attached both of them.
> 
> What's wrong is that this C code:
> 
> 	S_ISREG(mode)
> 
> is *not* equivalent to this C# code you use:
> 
> 	(stat.st_mode & MUN.FilePermissions.S_IFREG) ==
> 	   MUN.FilePermissions.S_IFREG
> 
> If you read /usr/include/sys/stat.h, S_ISREG is this:
> 
> 	#define __S_ISTYPE(mode, mask) (((mode) & __S_IFMT) == (mask))
> 	#define S_ISREG(mode) __S_ISTYPE((mode), __S_IFREG)
> 
> In short, S_ISREG is equivalent to this C# code:
> 
> 	(stat.st_mode & MUN.FilePermissions.S_IFMT) == 
> 	  MUN.FilePermissions.S_IFREG;
> 
> Ditto for all the other S_IS* macro calls.
> 
> So your C# code is buggy. :-)

That's apparent to me now. ;-)

> You might try looking at the internal method
> Mono.Unix.UnixFileSystemInfo.IsFileType(), which is used by the public
> properties IsDirectory, IsFifo, etc.

UnixFileSystemInfo seems to be a well-designed abstraction on top of the
Unix VFS API, but I'll stick with syscalls, because I don't like
UnixFileSystemInfo's exception policy.

Thanks for your (extremely clear and conscise) explanation.

-- 
Laci

    Blog: http://monda.hu/~laci/blog
    Home: http://mondalaci.objectis.net