// Consider initializing the field directly instead of declaring a .cctor. There is a perf
// tradeoff to either choice. The tradeoff is some startup perf if you compile code
// that references this type, vs. a bit of a hit every time you use one of these methods
// because we must verify that the .cctor ran. For now, we'll make this class slower
// to use, but you can write code with quirks in it that only triggers iff you need legacy
// behavior (ie, if you need to not throw an exception, see if you need to throw it first
// via normal logic then check the compatibility flag).
// Initialize the s_map field right before it's going to be used, not the moment we
// JITs code that contains a call to this class. For this reason, explicitly create a
// .cctor instead of letting the compiler generate one with different initialization
// semantics.
static BinaryCompatibility()
{
s_map = new BinaryCompatibilityMap();
}
// Consider initializing the field directly instead of declaring a .cctor. There is a perf
// tradeoff to either choice. The tradeoff is some startup perf if you compile code
// that references this type, vs. a bit of a hit every time you use one of these methods
// because we must verify that the .cctor ran. For now, we'll make this class slower
// to use, but you can write code with quirks in it that only triggers iff you need legacy
// behavior (ie, if you need to not throw an exception, see if you need to throw it first
// via normal logic then check the compatibility flag).
// Initialize the s_map field right before it's going to be used, not the moment we
// JITs code that contains a call to this class. For this reason, explicitly create a
// .cctor instead of letting the compiler generate one with different initialization
// semantics.
static BinaryCompatibility()
{
s_map = new BinaryCompatibilityMap();
}
