Re: core symlinks

[terjem () stud cs uit no (Terje Normann Marthinussen)]

> > Well, here's one data point. I just tested under SunOS 4.1.1; a core dump
> > will follow a symlink, unless the symlink points to a file that the user
> > can't write, in which case no core will be dumped. I think this is
> > reasonable behavior.
>
> I think it is too, and here's at least one legitimate use of core symlinks.
> We've had problems in the past with core dumps from our 32 Meg machines
> that come from daemon processes filling up the root file system, so we
> solved that problem by making /core a symlink to /dev/null.

I just came to think about another usefull purpose.
Like mentioned recently here, some login programs that reads the shadow
password file might place it, or parts of it, in its world read and writable 
core file if it crash. I believe this could be fixed by linking /core to 
someplace only accessable by root. 

Terje Marthinussen
terjem () stud cs uit no