Invalid function pointers crash Nidhugg

[Redhotsmasher]

Nidhugg does not detect when the program under test segfaults by attempting to call an invalid function pointer (see MWE attached below). I would expect Nidhugg to detect this and print an error about the program under test causing a segmentation fault (as it does when dereferencing an invalid non-function pointer) but this apparently does not work for function pointers causing Nidhugg itself to segfault.

segfaulttest.zip

[margnus1]

Thank you for your report, your example reproduces the problem nicely.

Because nidhugg is built on top of the llvm interpreter, and the llvm interpreter is not defensively implemented, there are unfortunately these kinds of behaviours in several places.

For this one, I guess we have to validate function pointers towards a list of known functions during indirect calls. I hope that's not hard to do.

[margnus1]

#112 Should fix the issue you're having, but if you don't mind I'll keep this issue open to track that all the general cases have not been addressed.

• Indirect call/pthread_create/atexit with null function pointer
• Indirect call/pthread_create/atexit with non-null invalid function pointer
• Indirect call/pthread_create/atexit with function pointer of wrong type (?)
• Safe reporting of execution of LLVM's "unreachable" instruction (can result if the program above is compiled with optimisations)