> A function that is called once will always be inlined.
That is objectively false. Even though I agree with the spirit. For starters, unoptimized builds won't have that. Second, exceptions tend to inhibit inlining. Compare https://godbolt.org/z/c8Jayf with https://godbolt.org/z/Uoo2YA. Third, it's easy enough to push one of the heuristics used for inlining high enough in the wrong direction (e.g. function size). I recommend the clang optimizer view on godbolt.
Also note that it can be hard for a compiler to prove that a function is not used outside the current translation unit (although anonymous namespaces help with that). Number of calls is only one of the heuristics for inlining though anyway.
That is objectively false. Even though I agree with the spirit. For starters, unoptimized builds won't have that. Second, exceptions tend to inhibit inlining. Compare https://godbolt.org/z/c8Jayf with https://godbolt.org/z/Uoo2YA. Third, it's easy enough to push one of the heuristics used for inlining high enough in the wrong direction (e.g. function size). I recommend the clang optimizer view on godbolt.
Also note that it can be hard for a compiler to prove that a function is not used outside the current translation unit (although anonymous namespaces help with that). Number of calls is only one of the heuristics for inlining though anyway.