[SYCL][TEST-E2E] Refactor the test to address Windows not printing th…

…e exception message (intel#14055)

- C++ thrown exception message not shown when running from Windows
terminal.
- The patch fixes
[cuda-max-local-mem-size.cpp](https://github.com/intel/llvm/blob/sycl/sycl/test-e2e/Plugin/cuda-max-local-mem-size.cpp)
test failure.

sycl/test-e2e/Plugin/cuda-max-local-mem-size.cpp

@@ -1,8 +1,8 @@
 // REQUIRES: cuda
 
 // RUN: %{build} -o %t.out
-// RUN: not %{run} SYCL_PI_CUDA_MAX_LOCAL_MEM_SIZE=0 %t.out 2>&1 | FileCheck --check-prefixes=CHECK-ZERO %s
-// RUN: not %{run} SYCL_PI_CUDA_MAX_LOCAL_MEM_SIZE=100000000 %t.out 2>&1 | FileCheck --check-prefixes=CHECK-OVERALLOCATE %s
+// RUN: %{run} SYCL_PI_CUDA_MAX_LOCAL_MEM_SIZE=0 %t.out 2>&1 | FileCheck --check-prefixes=CHECK-ZERO %s
+// RUN: %{run} SYCL_PI_CUDA_MAX_LOCAL_MEM_SIZE=100000000 %t.out 2>&1 | FileCheck --check-prefixes=CHECK-OVERALLOCATE %s
 
 //==---------------------- cuda-max-local-mem-size.cpp --------------------===//
 //==--- SYCL test to test SYCL_PI_CUDA_MAX_LOCAL_MEM_SIZE env var----------===//
@@ -16,15 +16,19 @@
 #include <sycl/detail/core.hpp>
 
 int main() {
-  sycl::queue Q{};
-  auto LocalSize =
-      Q.get_device().get_info<sycl::info::device::local_mem_size>();
-  Q.submit([&](sycl::handler &cgh) {
-     auto LocalAcc = sycl::local_accessor<float>(LocalSize + 1, cgh);
-     cgh.parallel_for(sycl::nd_range<1>{32, 32}, [=](sycl::nd_item<1> idx) {
-       LocalAcc[idx.get_global_linear_id()] *= 2;
-     });
-   }).wait();
+  try {
+    sycl::queue Q{};
+    auto LocalSize =
+        Q.get_device().get_info<sycl::info::device::local_mem_size>();
+    Q.submit([&](sycl::handler &cgh) {
+       auto LocalAcc = sycl::local_accessor<float>(LocalSize + 1, cgh);
+       cgh.parallel_for(sycl::nd_range<1>{32, 32}, [=](sycl::nd_item<1> idx) {
+         LocalAcc[idx.get_global_linear_id()] *= 2;
+       });
+     }).wait();
+  } catch (const std::exception &e) {
+    std::puts(e.what());
+  }
   // CHECK-ZERO: Local memory for kernel exceeds the amount requested using SYCL_PI_CUDA_MAX_LOCAL_MEM_SIZE
   // CHECK-OVERALLOCATE: Excessive allocation of local memory on the device
 }