Opencv4.6 compitible problem with opencv4.11

C++
1

Our program is developed under opencv 4.6, now the new environment is opencv 4.11 installed. It shows 127 error when i run the code.
I assume it is mixing library problem. how can i make the independent library 406 substituted library 411?

$ ldd xxxx | grep opencv
libopencv_aruco.so.406 => /usr/local/lib/libopencv_aruco.so.406 (0x0000ffffa0ab0000)
libopencv_objdetect.so.406 => /usr/local/lib/libopencv_objdetect.so.406 (0x0000ffffa0990000)
libopencv_calib3d.so.406 => /usr/local/lib/libopencv_calib3d.so.406 (0x0000ffffa0760000)
libopencv_imgcodecs.so.406 => /usr/local/lib/libopencv_imgcodecs.so.406 (0x0000ffffa0680000)
libopencv_features2d.so.406 => /usr/local/lib/libopencv_features2d.so.406 (0x0000ffffa05c0000)
libopencv_imgproc.so.406 => /usr/local/lib/libopencv_imgproc.so.406 (0x0000ffffa0140000)
libopencv_core.so.406 => /usr/local/lib/libopencv_core.so.406 (0x0000ffff9f5c0000)
libopencv_dnn.so.411 => /usr/local/lib/libopencv_dnn.so.411 (0x0000ffff9c8b0000)
libopencv_flann.so.411 => /usr/local/lib/libopencv_flann.so.411 (0x0000ffff9c820000)

2

yes, you cannot combine parts of opencv that have different versions. they should all be of the same version. if you do mix them, expect issues. I can’t guarantee that it’ll always fail but the chances of that going well are very slim.

how did you accomplish that linkage anyway? were there particular reasons to attempt this?

3

Dear Crackwitz, thank you for your attention.

  1. Yes, I did manually mixed link different versions of OpenCV libraries (4.6 and 4.11). This happens when:
  • The original program is compiled with OpenCV 4.6 to generate binaries (computeCalibration and applyCalibration)
  • After the system upgrade, I installed OpenCV 4.11 as the main development environment
  • When running the Python GUI (which depends on OpenCV 4.11), it loads a mixture of old and new shared libraries
  1. technical background:
    This mixed linking is achieved by:
  • Explicit library path specification at compile time (-L/path/to/opencv4.6/lib)
  • LD_LIBRARY_PATH environment variable conflict at runtime
  • Include path confusion caused by not thoroughly cleaning up old header files

3.motivation for upgrading:
The main reasons for my upgrade to OpenCV 4.11 are:

  1. Need 4.8+ version for GPU acceleration support for Jetson Orin Nano
  2. Fix the GStreamer memory leak problem in 4.6
  3. Use the DNN module optimization newly added in 4.11

Thank you

4

so… can you recompile the program? then the solution is simple. do that, against one version of opencv.