Running bundle adjustment ... Failed to compute a step: CHOLMOD warning: Matrix not positive definite.

[RigerLee]

ubuntu 20.04
ceres 2.1.0
eigen 3.4.0

Glomap doesnot work with gerrard-hall dataset without linking to OpenMP::OpenMP_CXX after commit 55e40b1 which updates parallel method from openmp to colmap threadpool.

Logs:
Running bundle adjustment ... I20241218 11:20:34.274386 2481332 global_mapper.cc:177] Bundle adjustment start W20241218 11:20:34.736182 2481332 levenberg_marquardt_strategy.cc:115] Linear solver failure. Failed to compute a step: CHOLMOD warning: Matrix not positive definite. W20241218 11:20:34.870784 2481332 levenberg_marquardt_strategy.cc:115] Linear solver failure. Failed to compute a step: CHOLMOD warning: Matrix not positive definite. W20241218 11:20:35.003857 2481332 levenberg_marquardt_strategy.cc:115] Linear solver failure. Failed to compute a step: CHOLMOD warning: Matrix not positive definite. W20241218 11:20:35.135973 2481332 levenberg_marquardt_strategy.cc:115] Linear solver failure. Failed to compute a step: CHOLMOD warning: Matrix not positive definite. W20241218 11:20:35.272040 2481332 levenberg_marquardt_strategy.cc:115] Linear solver failure. Failed to compute a step: CHOLMOD warning: Matrix not positive definite. E20241218 11:20:35.272112 2481332 trust_region_minimizer.cc:95] Terminating: Number of consecutive invalid steps more than Solver::Options::max_num_consecutive_invalid_steps: 5 I20241218 11:20:35.288867 2481332 bundle_adjustment.cc:48] Ceres Solver Report: Iterations: 5, Initial cost: 9.123054e+05, Final cost: 9.123054e+05, Termination: FAILURE

The result would be like below:

Using openmp version, it works:

Strange behavior.
Hope this will be helpful for those who encounter the same problem.

[StonerLing]

I encountered the same issue, but not with all datasets; it seems to happen with some super large datasets in my experiments.