Any interest in swat enhancement

Peter Samuelson peter at
Tue Sep 5 09:06:29 GMT 2000

[Ron Alexander <rcalex at>]
> I guess I will have to wait for someone to create the kill cmd. Just
> take my word that porting the Linux version of kill or even
> attempting to create one from scratch from POSIX documentation will
> not work.

The kill command itself is trivial.  The problem is that your OS has to
have support for the concept of sending and catching signals between
processes, and the concept of having many distinct signals.

This pretty much *has* to have significant kernel support, because
signals are handled asynchronously -- the process that gets one jumps
to a signal handler no matter what is going on at the time -- rather
like an interrupt handler at the kernel level.

For the case of controlling the debug level, it is not strictly
necessary to have the full semantics of signals.  One alternate
implementation: in the DEBUG() call, do a stat() on a special (not
literally) file; if the file has been changed since last time, open and
read it; it will contain the new debug level.  Then, to change the
debug level, you only have to update the debug-level file.


More information about the samba-technical mailing list