URDF relative paths that don't start with "./"

[stephane-caron]

The update from https://github.com/stack-of-tasks/pinocchio/pull/1742 addressed https://github.com/stack-of-tasks/pinocchio/issues/1741, but people are inventive and there is another relative-path syntax out there :wink:

There are currently 5 / 45 descriptions configured in robot_descriptions.py that use relative paths:

Description	Pinocchio 2.6.4	Pinocchio 2.6.10
cf2_description	❌	:heavy_check_mark:
laikago_description	❌	:heavy_check_mark:
mini_cheetah_description	❌	❌
minitaur_description	❌	❌
pr2_description	❌	❌

https://github.com/stack-of-tasks/pinocchio/pull/1742 solved descriptions that do <mesh filename="./foo/bar">, but it doesn't work on the remaining 3 because they do <mesh filename="foo/bar">.

[jcarpent]

@fabinsch COuld you handle it?

[stephane-caron]

I don't have a Pinocchio build environment to test it right now, but the fix may be simple.

In:

          if ( bf::exists( bf::path(package_dirs[i] + "/" + string.substr(2))))
          {
            result_path = std::string( package_dirs[i] + "/" + string.substr(2));
            break;
          }

Can the calls to substr(2) be removed? Substrings like "/./" (or "/./.././../") shouldn't hurt the path.

Also note that a URDF may do <mesh filename="../foo/bar">.

[fabinsch]

Hey @stephane-caron , thanks for raising the issue. You are right, and the solution that you're proposing works. I created a PR #1748 which should fix your loading issues in https://github.com/stephane-caron/robot_descriptions.py/issues/6.

[jcarpent]

@stephane-caron I have merged #1748. Could you check if it is now working well on your side?

[stephane-caron]

Could you check if it is now working well on your side?

Sorry this one went out of my radar. I've opened a follow up at https://github.com/stack-of-tasks/pinocchio/issues/1786