[OpenMP] Fix tests relying on the heap size variable

Summary:
I made that an unimplemented error, but forgot that it was used for this
environment variable.

Author: jhuber6

offload/plugins-nextgen/common/include/PluginInterface.h

@@ -1063,6 +1063,7 @@ struct GenericDeviceTy : public DeviceAllocatorTy {
 
   virtual Error getDeviceStackSize(uint64_t &V) = 0;
 
+  virtual bool hasDeviceHeapSize() { return false; }
   virtual Error getDeviceHeapSize(uint64_t &V) {
     return Plugin::error(error::ErrorCode::UNSUPPORTED,
                          "%s not supported by platform", __func__);

offload/plugins-nextgen/common/src/PluginInterface.cpp

@@ -762,13 +762,15 @@ Error GenericDeviceTy::init(GenericPluginTy &Plugin) {
     return StackSizeEnvarOrErr.takeError();
   OMPX_TargetStackSize = std::move(*StackSizeEnvarOrErr);
 
-  auto HeapSizeEnvarOrErr = UInt64Envar::create(
-      "LIBOMPTARGET_HEAP_SIZE",
-      [this](uint64_t &V) -> Error { return getDeviceHeapSize(V); },
-      [this](uint64_t V) -> Error { return setDeviceHeapSize(V); });
-  if (!HeapSizeEnvarOrErr)
-    return HeapSizeEnvarOrErr.takeError();
-  OMPX_TargetHeapSize = std::move(*HeapSizeEnvarOrErr);
+  if (hasDeviceHeapSize()) {
+    auto HeapSizeEnvarOrErr = UInt64Envar::create(
+        "LIBOMPTARGET_HEAP_SIZE",
+        [this](uint64_t &V) -> Error { return getDeviceHeapSize(V); },
+        [this](uint64_t V) -> Error { return setDeviceHeapSize(V); });
+    if (!HeapSizeEnvarOrErr)
+      return HeapSizeEnvarOrErr.takeError();
+    OMPX_TargetHeapSize = std::move(*HeapSizeEnvarOrErr);
+  }
 
   // Update the maximum number of teams and threads after the device
   // initialization sets the corresponding hardware limit.

offload/plugins-nextgen/cuda/src/rtl.cpp

@@ -1242,6 +1242,7 @@ struct CUDADeviceTy : public GenericDeviceTy {
   Error setDeviceStackSize(uint64_t Value) override {
     return setCtxLimit(CU_LIMIT_STACK_SIZE, Value);
   }
+  bool hasDeviceHeapSize() override { return true; }
   Error getDeviceHeapSize(uint64_t &Value) override {
     return getCtxLimit(CU_LIMIT_MALLOC_HEAP_SIZE, Value);
   }

offload/test/offloading/interop-print.c

@@ -8,6 +8,7 @@
 
 // REQUIRES: gpu
 // XFAIL: nvptx64-nvidia-cuda
+// XFAIL: nvptx64-nvidia-cuda-LTO
 
 #include <omp.h>
 #include <stdio.h>