2022-7-11 pcl double free or corruption(out) . valgrind. -march=native_pcl double free or corruption (out)-优快云博客

本文链接：https://blog.youkuaiyun.com/HITORANGE/article/details/125727364

这篇博客记录了在使用PCL（Point Cloud Library）进行三维点云处理时遇到的内存泄漏问题。通过Valgrind工具进行内存检查，发现了一个double free或corruption错误，涉及Eigen库的内存对齐问题。经过调整，问题定位到PCL的GreedyProjectionTriangulation对象在析构时导致的内存释放异常。通过修改代码，确保正确释放内存，解决了这个问题。最后，再次运行Valgrind确认已无内存泄漏。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

hit@hit:~/qt/pclsurface/build$ valgrind ./pclsurface 1
==5651== Memcheck, a memory error detector
==5651== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==5651== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info
==5651== Command: ./pclsurface 1
==5651== 
load point clouds 304
cloud point number : 304
kd tree established 
save surface.pcd 
==5651== Invalid free() / delete / delete[] / realloc()
==5651==    at 0x4C32D3B: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5651==    by 0x117155: Eigen::internal::aligned_free(void*) (Memory.h:177)
==5651==    by 0x1282D5: Eigen::aligned_allocator<Eigen::Matrix<float, 3, 1, 0, 3, 1> >::deallocate(Eigen::Matrix<float, 3, 1, 0, 3, 1>*, unsigned long) (Memory.h:747)
==5651==    by 0x12647F: std::allocator_traits<Eigen::aligned_allocator<Eigen::Matrix<float, 3, 1, 0, 3, 1> > >::deallocate(Eigen::aligned_allocator<Eigen::Matrix<float, 3, 1, 0, 3, 1> >&, Eigen::Matrix<float, 3, 1, 0, 3, 1>*, unsigned long) (alloc_traits.h:328)
==5651==    by 0x123A75: std::_Vector_base<Eigen::Matrix<float, 3, 1, 0, 3, 1>, Eigen::aligned_allocator<Eigen::Matrix<float, 3, 1, 0, 3, 1> > >::_M_deallocate(Eigen::Matrix<float, 3, 1, 0, 3, 1>*, unsigned long) (stl_vector.h:180)
==5651==    by 0x120015: std::_Vector_base<Eigen::Matrix<float, 3, 1, 0, 3, 1>, Eigen::aligned_allocator<Eigen::Matrix<float, 3, 1, 0, 3, 1> > >::~_Vector_base() (stl_vector.h:162)
==5651==    by 0x11C3E2: std::vector<Eigen::Matrix<float, 3, 1, 0, 3, 1>, Eigen::aligned_allocator<Eigen::Matrix<float, 3, 1, 0, 3, 1> > >::~vector() (stl_vector.h:435)
==5651==    by 0x133D69: pcl::GreedyProjectionTriangulation<pcl::PointNormal>::~GreedyProjectionTriangulation() (gp3.h:130)
==5651==    by 0x115189: main (main.cpp:77)
==5651==  Address 0x1f0fb6a0 is 32 bytes inside a block of size 3,680 alloc'd
==5651==    at 0x4C31B0F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5651==    by 0x975CE3C: std::vector<Eigen::Matrix<float, 3, 1, 0, 3, 1>, Eigen::aligned_allocator<Eigen::Matrix<float, 3, 1, 0, 3, 1> > >::reserve(unsigned long) (in /usr/local/lib/libpcl_surface.so.1.12.1.99)
==5651==    by 0x9767B83: pcl::GreedyProjectionTriangulation<pcl::PointNormal>::reconstructPolygons(std::vector<pcl::Vertices, std::allocator<pcl::Vertices> >&) (in /usr/local/lib/libpcl_surface.so.1.12.1.99)
==5651==    by 0x976BD9A: pcl::GreedyProjectionTriangulation<pcl::PointNormal>::performReconstruction(pcl::PolygonMesh&) (in /usr/local/lib/libpcl_surface.so.1.12.1.99)
==5651==    by 0x119DB1: pcl::MeshConstruction<pcl::PointNormal>::reconstruct(pcl::PolygonMesh&) (reconstruction.hpp:165)
==5651==    by 0x115140: main (main.cpp:96)
==5651== 
==5651== 
==5651== HEAP SUMMARY:
==5651==     in use at exit: 3,920 bytes in 4 blocks
==5651==   total heap usage: 46,386 allocs, 46,383 frees, 4,645,641 bytes allocated
==5651== 
==5651== LEAK SUMMARY:
==5651==    definitely lost: 3,680 bytes in 1 blocks
==5651==    indirectly lost: 0 bytes in 0 blocks
==5651==      possibly lost: 0 bytes in 0 blocks
==5651==    still reachable: 240 bytes in 3 blocks
==5651==         suppressed: 0 bytes in 0 blocks
==5651== Rerun with --leak-check=full to see details of leaked memory
==5651== 
==5651== For counts of detected and suppressed errors, rerun with: -v
==5651== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)

double free or corruption when :

    //初始化GreedyProjectionTriangulation对象，并设置参数
    pcl::GreedyProjectionTriangulation<pcl::PointNormal> gp3;
    //创建多变形网格，用于存储结果
    pcl::PolygonMesh triangles;

    //设置GreedyProjectionTriangulation对象的参数
    //第一个参数影响很大
    gp3.setSearchRadius(200.0f);           //设置连接点之间的最大距离（最大边长）用于确定k近邻的球半径【默认值 0】
    gp3.setMu(2.5f);                       //设置最近邻距离的乘子，以得到每个点的最终搜索半径【默认值 0】
    gp3.setMaximumNearestNeighbors(100);   //设置搜索的最近邻点的最大数量
    gp3.setMaximumSurfaceAngle(M_PI / 4);  // 45 degrees（pi）最大平面角
    gp3.setMinimumAngle(M_PI / 18);        // 10 degrees 每个三角的最小角度
    gp3.setMaximumAngle(2 * M_PI / 3);     // 120 degrees 每个三角的最大角度
    gp3.setNormalConsistency(false);       //如果法向量一致，设置为true

    //设置搜索方法和输入点云
    gp3.setInputCloud(cloud_with_normals);
    gp3.setSearchMethod(tree2);

//    //执行重构，结果保存在triangles中
    gp3.reconstruct(triangles);

problem localization using
valgrind :
valgrind introduction

Eigen !

Eigen memory align

Eigen字节对齐问题

视觉SLAM十四讲报错ch12: segmentation fault (core dumped) ./pointcloud_mapping

double free or corruption in pcl -march=native

solve :

ubuntu16.04安装点云库PCL

add : pcl cmake -march=native .. failed .

==15723== HEAP SUMMARY:
==15723==     in use at exit: 3,920 bytes in 4 blocks
==15723==   total heap usage: 46,386 allocs, 46,383 frees, 4,645,641 bytes allocated
==15723== 
==15723== LEAK SUMMARY:
==15723==    definitely lost: 3,680 bytes in 1 blocks
==15723==    indirectly lost: 0 bytes in 0 blocks
==15723==      possibly lost: 0 bytes in 0 blocks
==15723==    still reachable: 240 bytes in 3 blocks
==15723==         suppressed: 0 bytes in 0 blocks
==15723== Rerun with --leak-check=full to see details of leaked memory
==15723== 
==15723== For counts of detected and suppressed errors, rerun with: -v
==15723== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)

my own program in qt : .pro : QMAKE_CXXFLAGS += -march=native

==15723== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 0 from 0)
hit@hit:~/qt/pclsurface/build$ valgrind ./pclsurface 1
==15769== Memcheck, a memory error detector
==15769== Copyright (C) 2002-2017, and GNU GPL'd, by Julian Seward et al.
==15769== Using Valgrind-3.13.0 and LibVEX; rerun with -h for copyright info
==15769== Command: ./pclsurface 1
==15769== 
load point clouds 304
cloud point number : 304
kd tree established 
save surface.pcd 
==15769== 
==15769== HEAP SUMMARY:
==15769==     in use at exit: 240 bytes in 3 blocks
==15769==   total heap usage: 46,386 allocs, 46,383 frees, 4,645,833 bytes allocated
==15769== 
==15769== LEAK SUMMARY:
==15769==    definitely lost: 0 bytes in 0 blocks
==15769==    indirectly lost: 0 bytes in 0 blocks
==15769==      possibly lost: 0 bytes in 0 blocks
==15769==    still reachable: 240 bytes in 3 blocks
==15769==         suppressed: 0 bytes in 0 blocks
==15769== Rerun with --leak-check=full to see details of leaked memory
==15769== 
==15769== For counts of detected and suppressed errors, rerun with: -v
==15769== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)