Trajopt hangs indefinitely

[mkoval]

@jeking04 and are having Trajopt hang in PlanToConfiguration on HERB. These are the arguments that are getting passed to _Plan:

robot = <robot:GenericRobot - herb (6e82ad314d655dbc7aa357d85d1df71e)>
request = {'init_info': {'endpoint': [4.841931484861842, -1.2604496207661349, 0.0, 1.316917484156596, -0.1436371914612835, 0.47154291358025385, 0.016050195341527466], 'type': 'straight_line'}, 'costs': [{'params': {'coeffs': [1]}, 'type': 'joint_vel'}, {'params': {'coeffs': [20], 'dist_pen': [0.025]}, 'type': 'collision'}], 'constraints': [{'params': {'vals': [4.841931484861842, -1.2604496207661349, 0.0, 1.316917484156596, -0.1436371914612835, 0.47154291358025385, 0.016050195341527466]}, 'type': 'joint'}], 'basic_info': {'n_steps': 10, 'manip': 'active', 'start_fixed': True}}
traj_constraints = ()
goal_constraints = ()
traj_costs = ()
goal_costs = ()
interactive = False
constraint_threshold = 0.0001
kwargs = {'defer': False}

In some cases, planning succeeds if I press <Ctrl> + C a few times and prints this warning:

[openravepy_global.cpp:682 pyGetIntFromPy] failed to extract int

In other cases, the planner SEGFAULTs.

@psigen Any ideas?

[psigen]

I also noticed this, and interestingly, I noticed that it doesn't seem to happen if I revert prpy to https://github.com/personalrobotics/prpy/commit/60635d46b7acbceba61075d745673465e89c2f7f. Is this the case for you guys as well?

[mkoval]

Here's what git bisect shows me:

$ git bisect log
git bisect start
# bad: [ae15302c3a0225a45bca6db34d60f6e0e3d51ad7] Merge pull request #237 from personalrobotics/feature/perception_simtrack
git bisect bad ae15302c3a0225a45bca6db34d60f6e0e3d51ad7
# good: [60635d46b7acbceba61075d745673465e89c2f7f] Added Travis:CI badge.
git bisect good 60635d46b7acbceba61075d745673465e89c2f7f
# good: [3b8b2b796eed33cb3d29deda375ab284f5b3309d] Merge pull request #221 from personalrobotics/trajopt_tests
git bisect good 3b8b2b796eed33cb3d29deda375ab284f5b3309d
# good: [9413adbb0cb239ca2e581d0953b20dd8a443982e] Merge pull request #231 from athackst/bugfix/perception-module
git bisect good 9413adbb0cb239ca2e581d0953b20dd8a443982e
# bad: [804d2b70ad60aacddbf333ea142592ac017b0b0e] Merge pull request #234 from personalrobotics/fix/unittiming_vel
git bisect bad 804d2b70ad60aacddbf333ea142592ac017b0b0e
# good: [15621ccf13402f522edf44e15aa857507eb512aa] Merge pull request #236 from personalrobotics/fix/chomp_opttraj
git bisect good 15621ccf13402f522edf44e15aa857507eb512aa
# bad: [ef05bca06a38cc988cf0eeccb2b1d3d708bef77c] fix ComputeUnitTiming velocity bug, add unit test
git bisect bad ef05bca06a38cc988cf0eeccb2b1d3d708bef77c
# first bad commit: [ef05bca06a38cc988cf0eeccb2b1d3d708bef77c] fix ComputeUnitTiming velocity bug, add unit test

[cdellin]

Haha excellent, that's my commit (-:

personalrobotics/prpy@ef05bca06a38cc988cf0eeccb2b1d3d708bef77c

[mkoval]

Fixed. This is such an obscure bug... https://github.com/personalrobotics/prpy/pull/239.

[cdellin]

Thanks @mkoval!