Enable conversion warnings

.azp/test.yml

@@ -29,6 +29,7 @@ variables:
   ctest_stride: 1
   mpiexec_timeout: 3000
   suite_name: $(Agent.JobName)
+  cxxflags: "-Werror"
 
 jobs:
 - job: linux_cpu

CMake/hoomd/HOOMDCUDASetup.cmake

@@ -7,7 +7,7 @@ MARK_AS_ADVANCED(ALWAYS_USE_MANAGED_MEMORY)
 if (ENABLE_HIP)
     if (HIP_PLATFORM STREQUAL "nvcc")
         # supress warnings in random123
-        set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcudafe --diag_suppress=code_is_unreachable")
+        set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcudafe --diag_suppress=code_is_unreachable -Xcompiler=-Wall -Xcompiler=-Wconversion")
 
         # setup nvcc to build for all CUDA architectures. Allow user to modify the list if desired
         if (CMAKE_CUDA_COMPILER_VERSION VERSION_GREATER 8.99)

CMakeLists.txt

@@ -63,7 +63,7 @@ set(CMAKE_CUDA_STANDARD 14)
 
 # Enable compiler warnings on gcc and clang (common compilers used by developers)
 if(CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wno-unknown-pragmas")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wconversion -Wno-sign-conversion -Wno-unknown-pragmas -Wno-deprecated-declarations")
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall")
 endif()
 

hoomd.code-workspace

@@ -82,7 +82,8 @@
             "*.cuh": "cpp",
             "*.in": "cpp",
             "cfenv": "cpp",
-            "fft": "cpp"
+            "fft": "cpp",
+            "__nullptr": "cpp"
         },
         "cSpell.words": [
             "Frenkel",

hoomd/Autotuner.cc

@@ -290,7 +290,7 @@ unsigned int Autotuner::computeOptimalParameter()
                 float sum = 0.0f;
                 for (std::vector<float>::iterator it = v.begin(); it != v.end(); ++it)
                     sum += *it;
-                m_sample_median[i] = sum/v.size();
+                m_sample_median[i] = sum/float(v.size());
                 }
             else if (m_mode == mode_max)
                 {

hoomd/BondedGroupData.cc

@@ -64,12 +64,7 @@ BondedGroupData<group_size, Group, name, has_type_mapping>::BondedGroupData(
         // offer a default type mapping
         for (unsigned int i = 0; i < n_group_types; i++)
             {
-            char suffix[2];
-            suffix[0] = 'A' + i;
-            suffix[1] = '\0';
-
-            std::string type_name = std::string(name) + std::string(suffix);
-            m_type_mapping.push_back(type_name);
+            m_type_mapping.push_back(getDefaultTypeName(i));
             }
         }
 
@@ -854,7 +849,7 @@ void BondedGroupData<group_size, Group, name, has_type_mapping>::rebuildGPUTable
 
             // allocate scratch buffers
             CachedAllocator& alloc = m_exec_conf->getCachedAllocator();
-            unsigned int tmp_size = m_groups.size()*group_size;
+            size_t tmp_size = m_groups.size()*group_size;
             unsigned int nptl = m_pdata->getN()+m_pdata->getNGhosts();
             ScopedAllocation<unsigned int> d_scratch_g(alloc, tmp_size);
             ScopedAllocation<unsigned int> d_scratch_idx(alloc, tmp_size);
@@ -1094,7 +1089,7 @@ void BondedGroupData<group_size, Group, name, has_type_mapping>::moveParticleGro
 
         MPI_Status stat;
         MPI_Request req;
-        unsigned int num = send_groups.size();
+        unsigned int num = (unsigned int)send_groups.size();
 
         MPI_Isend(&num, 1, MPI_UNSIGNED, new_rank, 0, m_exec_conf->getMPICommunicator(), &req);
         MPI_Wait(&req, &stat);
@@ -1174,7 +1169,7 @@ void BondedGroupData<group_size, Group, name, has_type_mapping>::moveParticleGro
             if (! is_local)
                 {
                 // append to end of group data
-                unsigned int n = m_groups.size();
+                unsigned int n = (unsigned int)m_groups.size();
                 m_group_tag.push_back(tag);
                 m_groups.push_back(members);
                 m_group_typeval.push_back(typeval);

hoomd/BondedGroupData.cu

@@ -7,10 +7,13 @@
 #include "ParticleData.cuh"
 #include "BondedGroupData.cuh"
 
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wconversion"
 #include <hip/hip_runtime.h>
 #include <thrust/sort.h>
 #include <thrust/device_ptr.h>
 #include <thrust/execution_policy.h>
+#pragma GCC diagnostic pop
 
 /*! \file BondedGroupData.cu
     \brief Implements the helper functions (GPU version) for updating the GPU bonded group tables

hoomd/BondedGroupData.h

@@ -267,7 +267,7 @@ class BondedGroupData
          */
         void removeAllGhostGroups()
             {
-            unsigned int new_size = m_groups.size() - m_n_ghost;
+            unsigned int new_size = (unsigned int)m_groups.size() - m_n_ghost;
             reallocate(new_size);
             m_n_ghost = 0;
             }
@@ -279,7 +279,7 @@ class BondedGroupData
          */
         void addGhostGroups(unsigned int ngroup)
             {
-            unsigned int new_size = m_groups.size()+ngroup;
+            unsigned int new_size = (unsigned int)m_groups.size() + ngroup;
             reallocate(new_size);
             m_n_ghost += ngroup;
             }
@@ -300,7 +300,7 @@ class BondedGroupData
         //! Get the number of group types
         unsigned int getNTypes() const
             {
-            return m_type_mapping.size();
+            return (unsigned int)m_type_mapping.size();
             }
 
         //! Return name of this template
@@ -678,7 +678,7 @@ class BondedGroupData
         //! Resize internal tables
         /*! \param new_size New size of local group tables, new_size = n_local + n_ghost
          */
-        void reallocate(unsigned int new_size)
+        void reallocate(size_t new_size)
             {
             m_groups.resize(new_size);
             m_group_typeval.resize(new_size);

hoomd/BoxDim.h

@@ -299,25 +299,25 @@ BoxDim
             #ifdef __HIPCC__
             if (m_periodic.z)
                 {
-                Scalar img = rintf(w.z * m_Linv.z);
+                Scalar img = rint(w.z * m_Linv.z);
                 w.z -= L.z * img;
                 w.y -= L.z * m_yz * img;
                 w.x -= L.z * m_xz * img;
                 }
 
             if (m_periodic.y)
                 {
-                Scalar img = rintf(w.y * m_Linv.y);
+                Scalar img = rint(w.y * m_Linv.y);
                 w.y -= L.y * img;
                 w.x -= L.y * m_xy * img;
                 }
 
             if (m_periodic.x)
                 {
-                w.x -= L.x * rintf(w.x * m_Linv.x);
+                w.x -= L.x * rint(w.x * m_Linv.x);
                 }
             #else
-            // on the cpu, branches are faster than calling rintf
+            // on the cpu, branches are faster than calling rint
             if (m_periodic.z)
                 {
                 if (w.z >= m_hi.z)
@@ -338,13 +338,13 @@ BoxDim
                 {
                 if (w.y >= m_hi.y)
                     {
-                    int i = (w.y*m_Linv.y+Scalar(0.5));
+                    int i = int(w.y*m_Linv.y+Scalar(0.5));
                     w.y -= (Scalar)i*L.y;
                     w.x -= (Scalar)i*L.y * m_xy;
                     }
                 else if (w.y < m_lo.y)
                     {
-                    int i = (-w.y*m_Linv.y+Scalar(0.5));
+                    int i = int(-w.y*m_Linv.y+Scalar(0.5));
                     w.y += (Scalar)i*L.y;
                     w.x += (Scalar)i*L.y * m_xy;
                     }
@@ -354,12 +354,12 @@ BoxDim
                 {
                 if (w.x >= m_hi.x)
                     {
-                    int i = (w.x*m_Linv.x+Scalar(0.5));
+                    int i = int(w.x*m_Linv.x+Scalar(0.5));
                     w.x -= (Scalar)i*L.x;
                     }
                 else if (w.x < m_lo.x)
                     {
-                    int i = (-w.x*m_Linv.x+Scalar(0.5));
+                    int i = int(-w.x*m_Linv.x+Scalar(0.5));
                     w.x += (Scalar)i*L.x;
                     }
                 }

hoomd/CMakeLists.txt

@@ -75,6 +75,13 @@ set(_hoomd_sources Analyzer.cc
                    filter/export_filters.cc
                    )
 
+# ignore conversion warnings in external files
+if(CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+    set_source_files_properties(extern/kiss_fft.cc PROPERTIES COMPILE_FLAGS "-Wno-conversion -Wno-float-conversion")
+    set_source_files_properties(extern/vmdsock.cc PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+    set_source_files_properties(extern/imd.cc PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+endif()
+
 set(_hoomd_headers
     AABB.h
     AABBTree.h
@@ -190,6 +197,17 @@ endforeach(getar_src)
 # quiet some warnings locally on files we can't modify
 if(CMAKE_COMPILER_IS_GNUCXX OR "${CMAKE_CXX_COMPILER_ID}" MATCHES "Clang")
     set_source_files_properties("extern/libgetar/vogl_miniz/vogl_miniz.cpp" PROPERTIES COMPILE_FLAGS "-Wno-strict-aliasing")
+    set_source_files_properties(extern/libgetar/src/DirArchive.cpp PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+    set_source_files_properties(extern/libgetar/src/SqliteArchive.cpp PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+    set_source_files_properties(extern/libgetar/src/TarArchive.cpp PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+    set_source_files_properties(extern/libgetar/src/ZipArchive.cpp PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+    set_source_files_properties(extern/libgetar/src/lz4.c PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+    set_source_files_properties(extern/libgetar/miniz/miniz.cpp PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+    if (CMAKE_COMPILER_IS_GNUCXX)
+        set_source_files_properties(extern/libgetar/sqlite3/sqlite3.c PROPERTIES COMPILE_FLAGS "-Wno-conversion -Wno-return-local-addr")
+    else()
+        set_source_files_properties(extern/libgetar/sqlite3/sqlite3.c PROPERTIES COMPILE_FLAGS "-Wno-conversion")
+    endif()
 endif()
 
 if (ENABLE_HIP)
@@ -206,6 +224,11 @@ add_library(HOOMD::_hoomd ALIAS _hoomd)
 # add quick hull as its own library so that it's symbols can be public
 add_library (quickhull SHARED extern/quickhull/QuickHull.cpp)
 
+# ignore conversion warnings in quickull
+if(CMAKE_COMPILER_IS_GNUCXX OR CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+    target_compile_options(quickhull PRIVATE "-Wno-conversion;-Wno-pessimizing-move")
+endif()
+
 # link the library to its dependencies
 target_link_libraries(_hoomd PUBLIC pybind11::pybind11 quickhull Eigen3::Eigen)
 

hoomd/CachedAllocator.h

@@ -77,7 +77,7 @@ class __attribute__((visibility("default"))) CachedAllocator
          * \returns a pointer to the allocated buffer
          */
         template<typename T>
-        T *getTemporaryBuffer(unsigned int num_elements);
+        T *getTemporaryBuffer(size_t num_elements);
 
         // Specifically allocate a char* buffer
         char *allocate(std::ptrdiff_t num_bytes)
@@ -106,8 +106,8 @@ class __attribute__((visibility("default"))) CachedAllocator
 
         bool m_managed;  //! True if we use unified memory
 
-        unsigned int m_num_bytes_tot;
-        unsigned int m_max_cached_bytes;
+        size_t m_num_bytes_tot;
+        size_t m_max_cached_bytes;
         float m_cache_reltol;
 
         free_blocks_type m_free_blocks;
@@ -163,7 +163,7 @@ class ScopedAllocation
     {
     public:
         //! Copy constructor
-        ScopedAllocation(CachedAllocator& alloc, unsigned int num_elements);
+        ScopedAllocation(CachedAllocator& alloc, size_t num_elements);
 
         //! Destructor
         ~ScopedAllocation();
@@ -184,15 +184,15 @@ class ScopedAllocation
     };
 
 template<typename T>
-T* CachedAllocator::getTemporaryBuffer(unsigned int num_elements)
+T* CachedAllocator::getTemporaryBuffer(size_t num_elements)
     {
     std::ptrdiff_t num_bytes = sizeof(T)*num_elements;
     char *result = 0;
 
     // short-cut to avoid storing duplicate NULL ptrs in the map
     if (!num_bytes) return (T*)NULL;
 
-    unsigned int num_allocated_bytes = num_bytes;
+    size_t num_allocated_bytes = num_bytes;
 
     // search the cache for a free block
     free_blocks_type::iterator free_block = m_free_blocks.lower_bound(num_bytes);
@@ -264,7 +264,7 @@ T* CachedAllocator::getTemporaryBuffer(unsigned int num_elements)
 
 //! Constructor
 template<typename T>
-ScopedAllocation<T>::ScopedAllocation(CachedAllocator& alloc, unsigned int num_elements)
+ScopedAllocation<T>::ScopedAllocation(CachedAllocator& alloc, size_t num_elements)
     : m_alloc(alloc)
     {
     data = m_alloc.getTemporaryBuffer<T>(num_elements);

hoomd/CellListGPU.cu

@@ -6,8 +6,11 @@
 
 #include "CellListGPU.cuh"
 
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wconversion"
 #include <thrust/sort.h>
 #include <thrust/device_ptr.h>
+#pragma GCC diagnostic pop
 
 /*! \file CellListGPU.cu
     \brief Defines GPU kernel code for cell list generation on the GPU