Segmentation Fault (Core Dumped) sometimes but not always when installing packages

[amorris28]

I'm on Ubuntu 20.04.1 Versions:

mamba 0.24.0
conda 4.13.0

At least 75% of the time I try to install a package, I get this error:

Segmentation fault (core dumped)

But if I rerun the same command, it will occasionally work. So, essentially I just have to keep rerunning the same command over and over until it completes. It can fail during the Looking for... ['package'] step or after the Confirm changes: [Y/n] y when it actually starts installing. Here is an example command:

mamba install -n aes -c conda-forge -c bioconda -c r r-vegan

This also happens when I install packages when I create an environment, for example:

mamba create -n env -c conda-forge -c bioconda -c r r-base

I do not have this problem when I install packages using a config.yml like this:

mamba env create -f config.yml

That appears to work every time and so my current work around is to edit the config.yml and do this every time I want to install a package:

mamba env remove -n aes
mamba env create -f config.yml

Any suggestions of how to troubleshoot this?

[wolfv]

what are the contents of the config.yml file?

[amorris28]

config.yml:

name: aes
channels:
        - conda-forge
        - bioconda
        - defaults
        - R
dependencies:
        - r-base=4.1
        - r-devtools
        - r-tidyverse=1.3.1
        - r-vegan
        - r-corncob
        - r-ggpubr
        - r-ggthemes
        - r-kableExtra
        - r-captioner
        - r-flextable
        - r-ggtext
        - bioconductor-phyloseq
        - bioconductor-dada2
        - bioconductor-decontam
        - bioconductor-ancombc
        - bioconductor-aldex2

[amacati]

I observe a similar behavior when trying to install pytorch. Steps to reproduce:

mamba create -n test python=3.11
mamba activate test
mamba install -c pytorch pytorch

-> Segmentation fault (core dumped)

[jonashaag]

@amacati can you reproduce this in a Docker container?

[amacati]

Seems specific to the installation I'm currently using. When using the micromamba Docker image the steps result in the expected "Could not solve" error, so I wasn't able to reproduce this. Still, my local installation seems to be broken and throws a Segfault. Strangely enough, I have only encountered it with Python 3.11 environments so far.

[wolfv]

We had an issue like this for unsolvable environments with 1.4.0 or 1.4.1, but the latest release fixes it (1.4.2).

Are you on the latest mamba locally @amacati?

[amacati]

I was still on 1.4.1, updating to 1.4.2 fixed the issue. Thanks!