diff --git a/GPU/Common/GPUCommonDef.h b/GPU/Common/GPUCommonDef.h
index d9a5bdf92b6ac..ffe5551f02f1b 100644
--- a/GPU/Common/GPUCommonDef.h
+++ b/GPU/Common/GPUCommonDef.h
@@ -30,7 +30,7 @@
 //Some GPU configuration settings, must be included first
 #include "GPUCommonDefSettings.h"
 
-#if !(defined(__CLING__) || defined(__ROOTCLING__) || defined(G__ROOT)) // No GPU code for ROOT
+#if !defined(__CLING__) && !defined(G__ROOT) // No GPU code for ROOT
   #if defined(__CUDACC__) || defined(__OPENCL__) || defined(__HIPCC__) || defined(__OPENCL_HOST__)
     #define GPUCA_GPUCODE // Compiled by GPU compiler
   #endif
diff --git a/GPU/Common/GPUCommonMath.h b/GPU/Common/GPUCommonMath.h
index 01067ca032650..b448a30107dad 100644
--- a/GPU/Common/GPUCommonMath.h
+++ b/GPU/Common/GPUCommonMath.h
@@ -521,7 +521,7 @@ GPUdi() void GPUCommonMath::AtomicMinInternal(S* addr, T val)
 #endif // GPUCA_GPUCODE
 }
 
-#if (defined(__CUDACC__) || defined(__HIPCC__)) && !defined(G__ROOT)
+#if (defined(__CUDACC__) || defined(__HIPCC__)) && !defined(G__ROOT) && !defined(__CLING__)
 #define GPUCA_HAVE_ATOMIC_MINMAX_FLOAT
 template <>
 GPUdii() void GPUCommonMath::AtomicMaxInternal(GPUglobalref() GPUgeneric() GPUAtomic(float) * addr, float val)
diff --git a/GPU/GPUTracking/Standalone/Benchmark/standalone.cxx b/GPU/GPUTracking/Standalone/Benchmark/standalone.cxx
index 95997a30034c9..a624e1e55ed4b 100644
--- a/GPU/GPUTracking/Standalone/Benchmark/standalone.cxx
+++ b/GPU/GPUTracking/Standalone/Benchmark/standalone.cxx
@@ -223,6 +223,14 @@ int32_t ReadConfiguration(int argc, char** argv)
     configStandalone.rec.tpc.nWaysOuter = 1;
     configStandalone.rec.tpc.trackReferenceX = 83;
     configStandalone.proc.outputSharedClusterMap = 1;
+    configStandalone.proc.clearO2OutputFromGPU = 1;
+    configStandalone.QA.clusterRejectionHistograms = 1;
+    configStandalone.proc.tpcIncreasedMinClustersPerRow = 500000;
+    configStandalone.proc.ignoreNonFatalGPUErrors = 1;
+    // TODO: rundEdx=1
+    // GPU_proc.qcRunFraction=$TPC_TRACKING_QC_RUN_FRACTION;"
+    // [[ $CTFINPUT == 1 ]] && GPU_CONFIG_KEY+="GPU_proc.tpcInputWithClusterRejection=1;"
+    // double pipeline / rtc
   }
 
   if (configStandalone.outputcontrolmem) {
@@ -893,7 +901,7 @@ int32_t main(int argc, char** argv)
       double pipelineWalltime = 1.;
       if (configStandalone.proc.doublePipeline) {
         HighResTimer timerPipeline;
-        if (RunBenchmark(rec, chainTracking, 1, iEvent, &nTracksTotal, &nClustersTotal) || RunBenchmark(recPipeline, chainTrackingPipeline, 2, iEvent, &nTracksTotal, &nClustersTotal)) {
+        if (configStandalone.proc.debugLevel < 2 && (RunBenchmark(rec, chainTracking, 1, iEvent, &nTracksTotal, &nClustersTotal) || RunBenchmark(recPipeline, chainTrackingPipeline, 2, iEvent, &nTracksTotal, &nClustersTotal))) {
           goto breakrun;
         }
         auto pipeline1 = std::async(std::launch::async, RunBenchmark, rec, chainTracking, configStandalone.runs, iEvent, &nTracksTotal, &nClustersTotal, 0, &timerPipeline);
diff --git a/GPU/GPUTracking/Standalone/CMakeLists.txt b/GPU/GPUTracking/Standalone/CMakeLists.txt
index 0b37ce38b79f7..48fbd77c62786 100644
--- a/GPU/GPUTracking/Standalone/CMakeLists.txt
+++ b/GPU/GPUTracking/Standalone/CMakeLists.txt
@@ -65,7 +65,12 @@ if (GPUCA_BUILD_DEBUG_SANITIZE)
     set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -shared-libasan")
   endif()
 endif()
-string(APPEND CMAKE_CXX_FLAGS " -Wno-error -Wall -Wextra -Wshadow -Wno-unused-function -Wno-unused-parameter -Wno-unused-local-typedefs -Wno-unknown-pragmas -Wno-write-strings")
+if(GPUCA_CONFIG_WERROR)
+  string(APPEND CMAKE_CXX_FLAGS " -Werror -Wno-unknown-warning-option")
+else()
+  string(APPEND CMAKE_CXX_FLAGS " -Wno-error")
+endif()
+string(APPEND CMAKE_CXX_FLAGS " -Wall -Wextra -Wshadow -Wno-unused-function -Wno-unused-parameter -Wno-unused-local-typedefs -Wno-unknown-pragmas -Wno-write-strings")
 string(APPEND CMAKE_SHARED_LINKER_FLAGS " -rdynamic -Wl,--no-undefined")
 if(CMAKE_CXX_COMPILER MATCHES "clang\\+\\+")
   string(APPEND CMAKE_CXX_FLAGS " -Wno-vla-cxx-extension")
diff --git a/GPU/GPUTracking/Standalone/cmake/build.sh b/GPU/GPUTracking/Standalone/cmake/build.sh
index 9fe650fd30905..2698a5bad3950 100755
--- a/GPU/GPUTracking/Standalone/cmake/build.sh
+++ b/GPU/GPUTracking/Standalone/cmake/build.sh
@@ -7,13 +7,16 @@ set -e
 mkdir -p standalone/build
 pushd standalone/build
 cp $1/GPU/GPUTracking/Standalone/cmake/config.cmake .
-cat >> config.cmake << "EOF"
-set(ENABLE_CUDA 1)
-set(ENABLE_HIP 1)
-set(ENABLE_OPENCL 1)
-set(GPUCA_CONFIG_ONNX 1)
-set(GPUCA_BUILD_EVENT_DISPLAY 0)
+if [[ $GPUCA_STANDALONE_CI == 1 ]]; then
+  cat >> config.cmake << "EOF"
+  set(ENABLE_CUDA 1)
+  set(ENABLE_HIP 1)
+  set(ENABLE_OPENCL 1)
+  set(GPUCA_CONFIG_ONNX 1)
+  set(GPUCA_BUILD_EVENT_DISPLAY 0)
+  set(GPUCA_CONFIG_WERROR 1)
 EOF
+fi
 cmake -DCMAKE_INSTALL_PREFIX=../ $1/GPU/GPUTracking/Standalone
 make ${JOBS+-j $JOBS} install
 popd
diff --git a/GPU/GPUTracking/Standalone/cmake/config.cmake b/GPU/GPUTracking/Standalone/cmake/config.cmake
index 77ce9c9e77fca..88fe418d40e5b 100644
--- a/GPU/GPUTracking/Standalone/cmake/config.cmake
+++ b/GPU/GPUTracking/Standalone/cmake/config.cmake
@@ -38,4 +38,5 @@ set(CUDA_COMPUTETARGET "default")           # 86 89
 #set(GPUCA_RTC_NO_COMPILED_KERNELS 1)
 #set(GPUCA_KERNEL_RESOURCE_USAGE_VERBOSE 1)
 #set(GPUCA_CONFIG_COMPILER gcc)             # gcc / clang
+#set(GPUCA_CONFIG_WERROR 1)
 #add_definitions(-DGPUCA_GPU_DEBUG_PRINT)
diff --git a/GPU/TPCFastTransformation/Spline2D.h b/GPU/TPCFastTransformation/Spline2D.h
index 788816f73d6dd..19f9fc7b55b91 100644
--- a/GPU/TPCFastTransformation/Spline2D.h
+++ b/GPU/TPCFastTransformation/Spline2D.h
@@ -22,7 +22,7 @@
 #include "FlatObject.h"
 #include "GPUCommonDef.h"
 
-#if !defined(__ROOTCLING__) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
+#if !defined(__CLING__) && !defined(G__ROOT) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
 #include <Vc/Vc>
 #include <Vc/SimdArray>
 #endif
diff --git a/GPU/TPCFastTransformation/Spline2DSpec.h b/GPU/TPCFastTransformation/Spline2DSpec.h
index e7a9db8e6ae9c..b4d351e8d0407 100644
--- a/GPU/TPCFastTransformation/Spline2DSpec.h
+++ b/GPU/TPCFastTransformation/Spline2DSpec.h
@@ -22,7 +22,7 @@
 #include "GPUCommonDef.h"
 #include "SplineUtil.h"
 
-#if !defined(__ROOTCLING__) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
+#if !defined(__CLING__) && !defined(G__ROOT) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
 #include <Vc/Vc>
 #include <Vc/SimdArray>
 #endif
diff --git a/GPU/TPCFastTransformation/SplineSpec.h b/GPU/TPCFastTransformation/SplineSpec.h
index 1af427dee503b..744be943181db 100644
--- a/GPU/TPCFastTransformation/SplineSpec.h
+++ b/GPU/TPCFastTransformation/SplineSpec.h
@@ -22,7 +22,7 @@
 #include "GPUCommonDef.h"
 #include "SplineUtil.h"
 
-#if !defined(__ROOTCLING__) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
+#if !defined(__CLING__) && !defined(G__ROOT) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
 #include <Vc/Vc>
 #include <Vc/SimdArray>
 #endif
diff --git a/GPU/TPCFastTransformation/devtools/IrregularSpline2D3D.h b/GPU/TPCFastTransformation/devtools/IrregularSpline2D3D.h
index d60326ddf76ab..33d6874f40aae 100644
--- a/GPU/TPCFastTransformation/devtools/IrregularSpline2D3D.h
+++ b/GPU/TPCFastTransformation/devtools/IrregularSpline2D3D.h
@@ -21,7 +21,7 @@
 #include "FlatObject.h"
 #include "GPUCommonDef.h"
 
-#if !defined(__ROOTCLING__) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
+#if !defined(__CLING__) && !defined(G__ROOT) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
 #include <Vc/Vc>
 #include <Vc/SimdArray>
 #endif
@@ -329,7 +329,7 @@ GPUdi() void IrregularSpline2D3D::getSplineVec(const float* correctedData, float
   // Same as getSpline, but using vectorized calculation.
   // \param correctedData should be at least 128-bit aligned
 
-#if !defined(__ROOTCLING__) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
+#if !defined(__CLING__) && !defined(G__ROOT) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
   const IrregularSpline1D& gridU = getGridU();
   const IrregularSpline1D& gridV = getGridV();
   int32_t nu = gridU.getNumberOfKnots();
diff --git a/GPU/TPCFastTransformation/devtools/SemiregularSpline2D3D.h b/GPU/TPCFastTransformation/devtools/SemiregularSpline2D3D.h
index 954738fa74f1b..835a525239346 100644
--- a/GPU/TPCFastTransformation/devtools/SemiregularSpline2D3D.h
+++ b/GPU/TPCFastTransformation/devtools/SemiregularSpline2D3D.h
@@ -23,7 +23,7 @@
 #include "RegularSpline1D.h"
 #include "FlatObject.h"
 
-#if !defined(__ROOTCLING__) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
+#if !defined(__CLING__) && !defined(G__ROOT) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
 #include <Vc/Vc>
 #include <Vc/SimdArray>
 #endif
@@ -398,7 +398,7 @@ inline void SemiregularSpline2D3D::getSplineVec(const float* correctedData, floa
   // Same as getSpline, but using vectorized calculation.
   // \param correctedData should be at least 128-bit aligned
 
-#if !defined(__ROOTCLING__) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
+#if !defined(__CLING__) && !defined(G__ROOT) && !defined(GPUCA_GPUCODE) && !defined(GPUCA_NO_VC)
   //&& !defined(__CLING__)
   /*
     Idea: There are 16 knots important for (u, v).