layers: In GetInstanceProcAddr, intercept instance procs even if NULL instance

This problem showed up when running cube --validate and the order
of the layers was arranged so that the threading layer is NOT first.
It can also show up when running vktrace and the threading layer
is in any position, because the vktrace layer becomes first.

The symptom is generally a failure to create an instance when the
threading layer is enabled.

The fix is to check for the non-instance-based procs first by
calling layer_intercept_instance_proc before bailing if the instance
is NULL.

1 files changed, 4 insertions, 4 deletions

diff --git a/layers/threading.cpp b/layers/threading.cpp
index fd7a4159..fcdd38b1 100644
--- a/layers/threading.cpp
+++ b/layers/threading.cpp
@@ -286,15 +286,15 @@ VK_LAYER_EXPORT PFN_vkVoidFunction VKAPI_CALL vkGetInstanceProcAddr(VkInstance i
     PFN_vkVoidFunction addr;
     layer_data* my_data;
 
-    if (instance == VK_NULL_HANDLE) {
-        return NULL;
-    }
-
     addr = layer_intercept_instance_proc(funcName);
     if (addr) {
         return addr;
     }
 
+    if (instance == VK_NULL_HANDLE) {
+        return NULL;
+    }
+
     my_data = get_my_data_ptr(get_dispatch_key(instance), layer_data_map);
     addr = debug_report_get_instance_proc_addr(my_data->report_data, funcName);
     if (addr) {