search

humanoid-path-planner / hpp-rbprm

"Implementation of RB-PRM planner using hpp."
5 stars 7 forks source link

Test kinodynamic fails #35

Closed florent-lamiraux closed 5 years ago

florent-lamiraux commented 5 years ago

I am trying to understand why test kinodynamic fails on recent devel branches.

I ran the test in valgrind and got an "Invalid read of size 8" at the following line: https://github.com/humanoid-path-planner/hpp-rbprm/blob/0cdb2ee8320af0752a1face9c3e887735d360d1d/src/utils/algorithms.cc#L424

This line is indeed suspicious since iterator e runs until plan.end () is reached, but in this case, e+1 should be out of bound. Surprisingly, adding assert (e+1 != plan.end ()) did not make the program fail.

pFernbach commented 5 years ago

Thanks for catching this error, I do not understand why the assert never fail ...

What was your error with the kinodynamic test ? It may be unrelated to this. With the devel versions of hpp of two days ago it work on my computer.

florent-lamiraux commented 5 years ago

The test fails on gepgitlab. The difference of behavior might be due to some uninitialized variable. https://gepgitlab.laas.fr/humanoid-path-planner/hpp-doc/-/jobs/17304

pFernbach commented 5 years ago

https://gepgitlab.laas.fr/humanoid-path-planner/hpp-doc/pipelines/4098

PR https://github.com/humanoid-path-planner/hpp-rbprm/pull/36 seems to fix the issue with the kinodynamic test with the devel branch.

florent-lamiraux commented 5 years ago

Thank you.