Closed mmurooka closed 1 year ago
Thanks for reporting the issue. It was a subtle problem with robot's configuration loading as explained in https://github.com/jrl-umi3218/mc_rtc/commit/a96d61aaa223e672c33bbb2109541e0498a4d4bf description.
Thanks for the fix @gergondet I have confirmed that the problem no longer occurs with or without the https://github.com/isri-aist/MultiContactController/commit/45838c6528b8b73105b0cd35e018b1676bb07bf4 change.
The mc_rtc controller sometimes has a segmentation fault at startup, although I have not yet found a clear reproduction condition. The backtrace printed by gdb is as follows. Strangely, this happened when I added surfaces in the robot rsdf file, or when I added the links in the environment URDF model. This problem occurred with MultiContactController, but I do not think it is controller-dependent.
The terminal message at controller startup is as follows. Note that robot-specific configuration files are not placed in this controller.
There seems to be a problem accessing configuration in the following line.
https://github.com/jrl-umi3218/mc_rtc/blob/b24fc0a340a77a867a05ae3d5857592846cb95b0/src/mc_control/MCController.cpp#L253
The problem appears to be avoided by using the following patch to prevent the addition of empty robot configurations. However, my understanding is that there should be no problem even if empty configurations are added, so I don't know why the problem occurs.
https://github.com/jrl-umi3218/mc_rtc/compare/master...mmurooka:robot-config-debug