Pinocchio 3.2.0 not working in python 3.10, python process being killed

[XiangningLin]

Bug description

I'm using Mac OS 15.0, and I need to use Pinocchio 3.2.0 and python 3.10 in my project. I'm using conda virtual env.

I used

conda install pinocchio=3.2.0 -c conda-forge

and then used my python3 and try to import pinocchio and print its version. But my python process directly gets killed.

Expected behavior

It should import pinocchio successfully and print the correct version 3.2.0

Reproduction steps

Steps to reproduce the behavior:

1. create a conda env

conda create --name pinocchio_test

2. get inside the env

conda activate pinocchio_test

3. run conda to install pinocchio

conda install -c conda-forge pinocchio=3.2.0 -y

4. run python

python3 -c "import pinocchio; print(pinocchio.__version__)"

6. the python process is being killed

Screenshots

My system is not out of memory, and if I run other .py file without using pinocchio, it works just like the "hello world".

System

• OS: MacOS 15.0
• Pinocchio version: 3.2.0

[jcarpent]

I can reproduce the issue. Why do you need Pinocchio 3.2.0?
Could you test with Pinocchio 3.4.0 directly?

[XiangningLin]

I tested with 3.4.0 directly, it has the same issue.

[jcarpent]

The issue seems to come from a fresh conda environment. I've triggered a new build on conda: conda-forge/pinocchio-feedstock#140 🤞

[jorisv]

Hello @XiangningLin,

Can you give us the result of conda list and conda info in your pinocchio_test environment ?

[jcarpent]

One quick solution is to create a new environment from the attached environment file using:
conda create --file=env.txt

env.txt

[jcarpent]

@traversaro Are you aware of any recent issues on Conda on ARM64 architecture? Currently, when installing Pinocchio in a fresh environment, we have a segfault on mac OS X ARM64 systems.

[traversaro]

@traversaro Are you aware of any recent issues on Conda on ARM64 architecture? Currently, when installing Pinocchio in a fresh environment, we have a segfault on mac OS X ARM64 systems.

Not that I am aware of, sorry. Do you happen to a conda list in a old working environment (for example #2619 (comment)) and a new non-working environment so that we can do a diff and check if there is that points out to a problem. Also a backtrace of the python segfaulting could point us to the problematic dependency.

[cmastalli]

@jcarpent and @traversaro, one of my students noticed this recently, including Crocoddyl, as well. However, I did not entirely think it was an error from Conda.

[jorisv]

Here the conda list result for a pinocchio fresh install crash.txt and the env.txt send by @jcarpent work.txt

I tried to create a new environment with pinocchio==3.3.1=py312hf720c60_0 and python 3.12.8 but it's still crashing.

When I run the python3 -c "import pinocchio" in lldb, there is no crash. I have tried to generate a dump file without any success…

[olivier-roussel]

Looks like the update of urdfdom from 4.0.1-h922ef61_0 to 4.0.1-h1874252_1 causes the problem

[jorisv]

There is not so much difference between 4.0.1-h922ef61_0 and 4.0.1-h1874252_1 PR:

• use rattle-build instead of conda-build
• rerender:
  • compiler update 17 -> 18
  • osx update 12 -> 13

There is no problem on:

• linux-64
• osx-64

[olivier-roussel]

@traversaro Any idea of what could have caused that in conda-forge/urdfdom-feedstock#34 ?

[XiangningLin]

I’m able to solve the problem by using the previous urdfdom build.Thank you guys!

…

On Tue, Mar 18, 2025 at 09:31 Olivier Roussel ***@***.***> wrote: @traversaro <https://github.com/traversaro> Any idea of what could have caused that in conda-forge/urdfdom-feedstock#34 <conda-forge/urdfdom-feedstock#34> ? — Reply to this email directly, view it on GitHub <#2619 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BBK3QPOT23SU3NVDJ5QUTE32VBC63AVCNFSM6AAAAABZD6G6NGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZTHEZDSOBZGE> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: olivier-roussel]*olivier-roussel* left a comment (stack-of-tasks/pinocchio#2619) <#2619 (comment)> @traversaro <https://github.com/traversaro> Any idea of what could have caused that in conda-forge/urdfdom-feedstock#34 <conda-forge/urdfdom-feedstock#34> ? — Reply to this email directly, view it on GitHub <#2619 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BBK3QPOT23SU3NVDJ5QUTE32VBC63AVCNFSM6AAAAABZD6G6NGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZTHEZDSOBZGE> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[traversaro]

Sorry, I was able to come back to this only this evening, and I do not have a osx-arm64 machine to test on until next morning. I tried to reproduce the issue in GitHub Actions in https://github.com/traversaro/pinocchio-urdfdom-debugging, but python3 -c "import pinocchio; print(pinocchio.__version__)" always works even on a huge matrix of different versions of pinocchio, urdfdom and python (see https://github.com/traversaro/pinocchio-urdfdom-debugging/actions/runs/13932782523). I wonder what is the difference between all your machines and GitHub Actions? Perhaps the use of zsh? Are you installing pinocchio with conda, micromamba or pixi?

[XiangningLin]

I’m installing using conda environment and I’m using zsh as well. Thanks!

…

On Tue, Mar 18, 2025 at 13:23 Silvio Traversaro ***@***.***> wrote: Sorry, I was able to come back to this only this evening, and I do not have a osx-arm64 machine to test on until next morning. I tried to reproduce the issue in GitHub Actions in https://github.com/traversaro/pinocchio-urdfdom-debugging, but python3 -c "import pinocchio; print(pinocchio.__version__)" always works even on a huge matrix of different versions of pinocchio, urdfdom and python (see https://github.com/traversaro/pinocchio-urdfdom-debugging/actions/runs/13932712133). I wonder what is the difference between all your machines and GitHub Actions? Perhaps the use of zsh? Are you installing pinocchio with conda, micromamba or pixi? — Reply to this email directly, view it on GitHub <#2619 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BBK3QPOVBBXIFNEHTSPZKGD2VB6DFAVCNFSM6AAAAABZD6G6NGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZUGYZDINBTGQ> . You are receiving this because you were mentioned.Message ID: ***@***.***> [image: traversaro]*traversaro* left a comment (stack-of-tasks/pinocchio#2619) <#2619 (comment)> Sorry, I was able to come back to this only this evening, and I do not have a osx-arm64 machine to test on until next morning. I tried to reproduce the issue in GitHub Actions in https://github.com/traversaro/pinocchio-urdfdom-debugging, but python3 -c "import pinocchio; print(pinocchio.__version__)" always works even on a huge matrix of different versions of pinocchio, urdfdom and python (see https://github.com/traversaro/pinocchio-urdfdom-debugging/actions/runs/13932712133). I wonder what is the difference between all your machines and GitHub Actions? Perhaps the use of zsh? Are you installing pinocchio with conda, micromamba or pixi? — Reply to this email directly, view it on GitHub <#2619 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BBK3QPOVBBXIFNEHTSPZKGD2VB6DFAVCNFSM6AAAAABZD6G6NGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDOMZUGYZDINBTGQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>