7.9. Dynamic Stack Tracing

Runtime safety checks are enforced by consulting information about the application call sequences (one per thread) and the call site for every call in these sequences. This information is called the context-sensitive call stack information. This information is available at any point during the execution and is maintained using a dynamic stack tracing mechanism. Programs are instrumented to efficiently and dynamically maintain their stack state at a high-level (source-code) and offer this information to the dynamic software updating runtime environment to enforce safety checks before an update is applied. The captured state is architecture (and operating system) independent.

The stack trace dynamically captures the names of functions that are active on the stack. For each function that is active on the stack, the instrumentation also saves the execution point from which the next stack frame was created when the callee function was called. The combination of function names and their execution points provides an accurate context-sensitive call stack trace. The execution points captured are equivalent to the continuation points described in Section 5.3.2. Using this call stack trace, safety checks such as type-safety and transaction-safety can be enforced more accurately. Type-safety can be enforced if type information is precomputed (statically) for every continuation point. Transaction-safety can be enforced if a user forbids updates from being applied inside specific regions of code which are active on the stack.