DISTRIBUTED BREAKPOINT DETECTION IN MESSAGE-PASSING PROGRAMS

The ability to set breakpoints stands, along with the possibility of d eterministic reexecution, as one of the most important issues in the d ebugging of message-passing programs. We consider in this paper the de sign of fully distributed algorithms for the detection of breakpoints in such programs, and provide four algorithms, one for each different type of breakpoint. One of the algorithms detects the occurrence of un conditional breakpoints, while the other three detect the occurrence o f breakpoints on disjunctive predicates, stable conjunctive predicates , and generic conjunctive predicates. All the algorithms we present de tect breakpoints in the form of earliest global states with respect to the particular property involved. In the case of unconditional breakp oints, such an earliest global state must coincide exactly with the re quested local unconditional breakpoints for the processes that do actu ally participate in the breakpoint. In the case of the other (conditio nal) breakpoints, what is detected is the earliest global state at whi ch either the disjunctive or the conjunctive predicate under considera tion is true. In order to actually halt the computation at the exact g lobal state the algorithms detect, we suggest as a first approach the use of checkpointing and rollback-recovery techniques. (C) 1996 Academ ic Press, Inc.