Re: problems with chmod-ing socket

Albert Dorofeev (
Thu, 18 Nov 1999 10:16:56 +0100

On Wed, Nov 17, 1999 at 11:30:02AM -0500, Michael Stenner wrote:
> On Wed, Nov 17, 1999 at 02:41:04PM +0100, Albert Dorofeev wrote:
> > On Wed, Nov 17, 1999 at 07:41:50AM -0500, Michael Stenner wrote:
> > > Here in the Duke Physics Dept, we had to change the "chmod 0700" in
> > > src/afterstep/modules.c to "chmod 4700" in order to get it to work.
> > > 
> > > We are using sun's NFS, which may be the source of our problems.
> > 
> > I would highly recommend that you change the location
> > of the pipe to the local filesystem. I did not have any
> > problems running it over NFS though but I can imagine
> > there could be. Just place the pipe onto the local disk.
> Hmm.. If that is the case, then perhaps it should be the default, or
> at least a compile time option.  
> I guess what I'm saying is this:
> this is a bug that should be fixed, either by afterstep, or sun.  I
> know of at least a couple of ways to do it in the afterstep
> code. (4700 or put it on a local drive)  We have implemented the

The mode of the socket was changed on my request. This is because
on some versions of Solaris the default mode for the pipe is
4777 despite the umask which allows anybody to send any commands
to AfterStep. The pipe's mode was changed right after its 
creation to minimize the risk. It is unfortunate chmod does
not take parameters like go-rwx so we have to take care of which
mask should be used. Apparently, it should be the one you say ...

Albert Dorofeev (Tigr) - albert at -
PGP fingerprint = C9 49 D0 F3 41 FA 8C D8  E9 5C 6A D4 F1 6D 65 15
  Anything good in life is either illegal, immoral or fattening.