What counts as automatic program analysis?

I'm trying to understand the definition of program analysis on Wikipedia:


Program analysis is the process of automatically analyzing the behavior of computer programs regarding a property such as correctness, robustness, safety and liveness.

What is the meaning of "automatic" here?

Are parsers static analysis tools, since they will throw an error if they encounter syntactically invalid code?

Is running a program an example of dynamic analysis, since it will provide me with an output (though I still have to interpret it myself)?

What if I explicitly throw a this can't happen exception when the program reaches an invalid state?

What about using a debugger to pause execution?

There is no sharp formal definition.

"Static" denotes "before running the program".

"Dynamic" denotes "while running the program".

"Automatic" stands for "without the intervention of the user". For instance, Hoare logic can be used to prove program properties, but (at least in its basic form) it requires user-provided invariants. Abstract interpretation, instead, often results in a totally automatic analysis. Type checking is usually regarded as automatic, since the type checker does not (usually) ask the user to provide hints about, e.g., why a given expression has the given type.

Parsing is static, but it's such a preliminary stage that probably most people wouldn't consider as program analysis at all. If a text file does not respect the syntactic rules, it is not part of the programming language (a language being a set of strings, after all!), hence it is not even a program, hence parsing is not program analysis. I would not be very interested in the syntax errors reported by cc kitten.jpeg, for instance.

Observing the program when it is being run is definitely a "dynamic" approach. Using a debugger, observing fired exceptions fall in this category.

StackExchange Computer Science Q#75837, answer score: 5