segfault address 0x1, cause 'memory not mapped'

[arronlacey]

Bug description

I am trying to rebuild the vetiver website within a Docker container. Running individual python chunks within the container work fine, but when trying to render the .qmd files I get the following:

$ quarto render index.qmd --to html


processing file: index.qmd
  |..............                                                        |  20%
  ordinary text without R code

  |............................                                          |  40%
label: unnamed-chunk-1
  |..........................................                            |  60%
  ordinary text without R code

  |........................................................              |  80%
label: unnamed-chunk-2 (with options) 
List of 1
 $ engine: chr "python"


 *** caught segfault ***
address 0x1, cause 'memory not mapped'

Traceback:
 1: py_call_impl(callable, dots$args, dots$keywords)
 2: builtins$eval(compiled, globals, locals)
 3: force(expr)
 4: py_capture_output(builtins$eval(compiled, globals, locals))
 5: py_compile_eval(snippet, compile_mode)
 6: reticulate::eng_python(options)
 7: engine(options)
 8: in_dir(input_dir(), expr)
 9: in_input_dir(engine(options))
10: block_exec(params)
11: call_block(x)
12: process_group.block(group)
13: process_group(group)
14: withCallingHandlers(if (tangle) process_tangle(group) else process_group(group),     error = function(e) {        setwd(wd)        cat(res, sep = "\n", file = output %n% "")        message("Quitting from lines ", paste(current_lines(i),             collapse = "-"), " (", knit_concord$get("infile"),             ") ")    })
15: process_file(text, output)
16: knitr::knit(knit_input, knit_output, envir = envir, quiet = quiet)
17: rmarkdown::render(input = input, output_format = output_format,     knit_root_dir = knit_root_dir, params = params, run_pandoc = FALSE,     envir = globalenv())
18: execute(params$input, params$format, params$tempDir, params$libDir,     params$dependencies, params$cwd, params$params, params$resourceDir,     params$handledLanguages, params$markdown)
19: .main()
An irrecoverable exception occurred. R is aborting now ...
root@d8296102c061:/home/conda/r_miniconda_container# 

I removed all python chunks from index.qmd to see if the error is raised, but it runs fine and then it fails in the next python chunk - the first evaluated python chunk in getting_started/index.qmd:

from vetiver.data import mtcars
from sklearn import linear_model

car_mod = linear_model.LinearRegression().fit(mtcars.drop(columns="mpg"), mtcars["mpg"])

Steps to reproduce:

docker pull registry.gitlab.com/arronlacey/r_miniconda_container
docker run --rm -p 8787:8787 registry.gitlab.com/arronlacey/r_miniconda_container
docker exec -it <container name> /bin/bash
git clone https://gitlab.com/arronlacey/r_miniconda_container.git
cd r_miniconda_container
quarto render index.qmd --to html

Dockerfile:

FROM rocker/r-ver:4.2.1

RUN apt-get update
RUN apt-get install -y --no-install-recommends libxt6
RUN apt-get install nano

ENV S6_VERSION=v2.1.0.2
ENV RSTUDIO_VERSION=2022.07.2+576
ENV DEFAULT_USER=rstudio
ENV PANDOC_VERSION=default
ENV QUARTO_VERSION=default

RUN /rocker_scripts/install_rstudio.sh
RUN /rocker_scripts/install_pandoc.sh
RUN /rocker_scripts/install_quarto.sh
RUN quarto install extension --no-prompt quarto-ext/grouped-tabsets

RUN useradd -u 1001 -g rstudio -m conda
USER conda

WORKDIR /home/conda

ENV PATH="/home/conda/miniconda3/bin:${PATH}"
ARG PATH="/home/conda/miniconda3/bin:${PATH}"
ENV MLFLOW_BIN=/home/conda/miniconda3/bin/mlflow
ENV MLFLOW_PYTHON_BIN=/home/conda/miniconda3/bin/python

RUN wget \
    https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh \
    && bash Miniconda3-latest-Linux-x86_64.sh -b \
    && rm -f Miniconda3-latest-Linux-x86_64.sh
RUN conda --version

USER root

RUN R -q -e 'install.packages(c("reticulate","vetiver","pins","dplyr","checkmate","quarto","gt","tidymodels","randomForest"))'
ENV PATH "$PATH:/root/.local/share/r-miniconda/bin"
RUN R -q -e 'reticulate::py_install(envname="base",c("vetiver","pins","scikit-learn","pandas","nbformat","nbclient"),pip=TRUE)'
ENV USER="rstudio"

CMD ["/usr/lib/rstudio-server/bin/rserver", "--server-daemonize", "0", "--auth-none", "1"]

EXPOSE 8787

CMD ["/init"]

Host details:

Distributor ID:	Ubuntu
Description:	Ubuntu 22.04.1 LTS
Release:	22.04
Codename:	jammy

Checklist

• Please include a minimal, fully reproducible example in a single .qmd file? Please provide the whole file rather than the snippet you believe is causing the issue.
• Please format your issue so it is easier for us to read the bug report.
• Please document the RStudio IDE version you're running (if applicable), by providing the value displayed in the "About RStudio" main menu dialog?
• Please document the operating system you're running. If on Linux, please provide the specific distribution.

[arronlacey]

interestingly some files will render that include python chunks - i.e. get_started/deploy.qmd but some files don't i.e. get_started/version.qmd.

an offending chunk in get_started/version.qmd is:

#| code-fold: true
#| code-summary: "Show the code from previous steps"
from vetiver.data import mtcars
from vetiver import VetiverModel
from sklearn import linear_model

car_mod = linear_model.LinearRegression().fit(mtcars.drop(columns="mpg"), mtcars["mpg"])

v = VetiverModel(car_mod, model_name = "cars_mpg",
                 save_ptype = True, ptype_data = mtcars)

but this chunk runs fine in deploy.qmd:

#| code-fold: true
#| code-summary: "Show the code from previous steps"
#| output: false

from vetiver.data import mtcars
from vetiver import VetiverModel, vetiver_pin_write
from sklearn import tree
from pins import board_folder

car_mod = tree.DecisionTreeRegressor().fit(mtcars.drop(columns="mpg"), mtcars["mpg"])

v = VetiverModel(car_mod, model_name = "cars_mpg",
                 save_ptype = True, ptype_data = mtcars.drop(columns="mpg"))

model_board = board_folder("pins-py", allow_pickle_read=True)
vetiver_pin_write(model_board, v)

so I'm really at a loss here...

[arronlacey]

I've isolated this line giving an error when rendering with quarto but works fine when running from a python shell:

car_mod = linear_model.LinearRegression().fit(mtcars.drop(columns="mpg"), mtcars["mpg"])

I only have one python installation in this container so I don't think it's a python issue, rather a quarto_cli issue

[cscheid]

I only have one python installation in this container so I don't think it's a python issue, rather a quarto_cli issue

The stack trace you have above shows the segfault happening inside reticulate:

Traceback:
 1: py_call_impl(callable, dots$args, dots$keywords)
 2: builtins$eval(compiled, globals, locals)
 3: force(expr)
 4: py_capture_output(builtins$eval(compiled, globals, locals))
 5: py_compile_eval(snippet, compile_mode)
 6: reticulate::eng_python(options)
...

For the sake of clarification: quarto-cli has no code that uses memory-unsafe languages, so a segfault almost by definition will be coming from somewhere else.

My guess is that the container you're using is somehow misconfiguring some option (or installation) in reticulate, which fails to load dynamically a library, and then has a null pointer instead of a function entry point.

[eitsupi]

Hey, isn't this the issue? rocker-org/rocker-versioned2#471 (comment)
This is tracking on the numpy repo. numpy/numpy#21643
quarto has nothing to do with this issue.

The workaround on rocker/r-ver is to switch the BLAS used in R.
https://rocker-project.org/images/versioned/r-ver.html#switching-blas-implementations

[arronlacey]

@eitsupi I think you are correct - so not a quarto issue.