Hi John! :-)

 JG> Hmmm, now that'd be an excellent security hole, if it didn't.... ;-)  
 JG> Picture a program setting a breakpoint and either monitoring the
 JG> caching of the hard drive till it picked up on the password file,
 JG> or found where the kernel was in memory, and just patched it to allow
 JG> its self to gain access to the entire system.... 
A user process could then monitor *when* a specific location is accessed,
but it could not read the contents.

Anyway, I was wrong in my original statement. If the debug registers were
not saved across task switches, a breakpoint triggering inside another
process would be delivered to that same process. That process would then
probably just crash because it doesn't expect to have to handle a
breakpoint (well, I'm not exactly sure how a debugging breakpoint is handed
over to the program since it can't have interrupt handlers of its own --
probably via a signal).

Ciao
Pascal

--- Msged/LNX 6.1.1