Fix uninitialized build warning

[MaximSmolskiy]

-- The C compiler identification is GNU 11.4.0
-- The CXX compiler identification is GNU 11.4.0

In file included from /home/mithridat/Downloads/pcl/recognition/src/implicit_shape_model.cpp:43:
/home/mithridat/Downloads/pcl/recognition/include/pcl/recognition/impl/implicit_shape_model.hpp: In member function ‘void pcl::features::ISMVoteList<PointT>::findStrongestPeaks(std::vector<pcl::ISMPeak, Eigen::aligned_allocator<pcl::ISMPeak> >&, int, double, double) [with PointT = pcl::PointXYZ]’:
/home/mithridat/Downloads/pcl/recognition/include/pcl/recognition/impl/implicit_shape_model.hpp:174:21: warning: ‘*(float*)((char*)&strongest_peak + offsetof(Eigen::Vector3f, Eigen::Matrix<float, 3, 1, 0, 3, 1>::<unnamed>.Eigen::PlainObjectBase<Eigen::Matrix<float, 3, 1, 0, 3, 1> >::<unnamed>.Eigen::MatrixBase<Eigen::Matrix<float, 3, 1, 0, 3, 1> >::<unnamed>.Eigen::DenseBase<Eigen::Matrix<float, 3, 1, 0, 3, 1> >::<unnamed>.Eigen::DenseCoeffsBase<Eigen::Matrix<float, 3, 1, 0, 3, 1>, 3>::<unnamed>.Eigen::DenseCoeffsBase<Eigen::Matrix<float, 3, 1, 0, 3, 1>, 1>::<unnamed>.Eigen::DenseCoeffsBase<Eigen::Matrix<float, 3, 1, 0, 3, 1>, 0>::<unnamed>))’ may be used uninitialized [-Wmaybe-uninitialized]
  174 |     Eigen::Vector3f strongest_peak;
      |                     ^~~~~~~~~~~~~~
/home/mithridat/Downloads/pcl/recognition/include/pcl/recognition/impl/implicit_shape_model.hpp:174:21: warning: ‘strongest_peak.Eigen::Matrix<float, 3, 1, 0, 3, 1>::<unnamed>.Eigen::PlainObjectBase<Eigen::Matrix<float, 3, 1, 0, 3, 1> >::m_storage.Eigen::DenseStorage<float, 3, 3, 1, 0>::m_data.Eigen::internal::plain_array<float, 3, 0, 0>::array[1]’ may be used uninitialized [-Wmaybe-uninitialized]

[mvieth]

Thank you for your pull request. I am pretty sure that the warning is a false positive: strongest_peak is only read from if peak_counter is not zero (line 191), which can only happen if strongest_peak was set to something valid (lines 185 and 187). I am assuming you are testing with PCL 1.14.0 or newer, see 053995b
I was also not able to reproduce the warning with GCC 11.5.0, so perhaps the maybe-uninitialized detection logic has been improved between the two compiler versions?
If you would still like to initialize strongest_peak with -1, please add an additional check in line 191, testing that strongest_peak is now different from -1 (to have a consistent logic)

[MaximSmolskiy]

Thank you for your pull request. I am pretty sure that the warning is a false positive: strongest_peak is only read from if peak_counter is not zero (line 191), which can only happen if strongest_peak was set to something valid (lines 185 and 187).

Yes, I also think that warning is false positive. And I think it's better to get rid of such warnings if it's not too difficult

I am assuming you are testing with PCL 1.14.0 or newer, see 053995b

Yes, I tested master branch

If you would still like to initialize strongest_peak with -1, please add an additional check in line 191, testing that strongest_peak is now different from -1 (to have a consistent logic)

There are four related variables - best_density, strongest_peak, best_peak_ind and peak_counter.
So, it's better to add additional checks for all of them in line 191 or leave as is - only for peak_counter?

[mvieth]

There are four related variables - best_density, strongest_peak, best_peak_ind and peak_counter. So, it's better to add additional checks for all of them in line 191 or leave as is - only for peak_counter?

I would at least like a check for strongest_peak, e.g. something like if( peak_counter == 0 || (strongest_peak(0) == -1.0f && strongest_peak(1) == -1.0f && strongest_peak(2) == -1.0f)) break;
You are right, a similar argument could be made for best_density and best_peak_ind, so feel free to add similar checks for them too, if you like.

[mvieth]

Thank you, looks good now.