You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Fix remaining failing test issues in current testbed (#13)
- Fix crashing issue where wrong instantiation parameter is passed
- Instead use the CORINFO_CALL_CODE_POINTER infrastructure to simply use an exact function pointer
- Additionally, my previous attempt to fix a different bug added the concept of allowing instantiation parameters to be required when using resolving the static virtual at compile time. In order to reliably use the CORINFO_CALL_CODE_POINTER infrastructure, that logic was deleted and removed.
- Fix issue where the type resolved via ResolveVirtualStaticMethod was on a slighly different interface type
- Fix by passing the interface type pointer to ResolveVirtualStaticMethod
- And by performing exact interface type checks in TryResolveVirtualStaticMethodOnThisType
- Finally there was 1 remaining bug in the test generator for the GenericContextTest where the correct string to test for was not generated
- Fix by deleting a hack used earlier to get closer to passing
- Drive by feature to verify that the MethodBody on a MethodImpl follows the rules around being a MethodDef and being implemented by the type that defines the MethodImpl.
// Check that there is no implementation of the interface on this type which is the canonical interface for a shared generic. If so, that indicates that
9181
+
// we cannot exactly compute a target method result, as even if there is an exact match in the type hierarchy
9182
+
// it isn't guaranteed that we will always find the right result, as we may find a match on a base type when we should find the match
0 commit comments