Question

When, if ever, is it safe to wrap behind a conditional code that requires an assembly that may not be present?

For example:

if (SafeCheckForOptionalAssembly()) {
    // Code requiring an optional assembly
} else {
    // fallback: don't rely on optional assembly
}

Directly related questions:

1. Does it matter if the conditionally included code uses a type contained within the optional assembly? (e.g. var foo = new MyClassInOptionalAssembly())
2. Does the behavior change if the conditionally included code is wrapped in a method?

Answer 1

1. Yes, doesn’t work. The jitter will bomb trying to compile the method before it starts running. The exception is raised in the calling method.

2. Possibly. But to be 100% sure, you’ll need to apply the [MethodImpl(MethodImplOptions.NoInlining)] attribute so the method never gets inlined.

These assumptions are only valid for a jitter that compiles on demand. The Microsoft x86 and x64 jitters do. The Mono jitter doesn’t (last I looked).