Setting up a custom image

I am trying to set up a custom docker image (linux-anvil-cos7-x86_64) but regenerate doesn’t take it into account.

conda-forge.yml:

docker:
  fallback_image: linux-anvil-cos7-x86_64

After regenerate, ./scripts/run_docker_build.sh contains:

if [ -z "${DOCKER_IMAGE}" ]; then
    SHYAML_INSTALLED="$(shyaml -h || echo NO)"
    if [ "${SHYAML_INSTALLED}" == "NO" ]; then
        echo "WARNING: DOCKER_IMAGE variable not set and shyaml not installed. Falling back to linux-anvil-cos7-x86_64"
        DOCKER_IMAGE="linux-anvil-cos7-x86_64"
    else
        DOCKER_IMAGE="$(cat "${FEEDSTOCK_ROOT}/.ci_support/${CONFIG}.yaml" | shyaml get-value docker_image.0 linux-anvil-cos7-x86_64 )"
    fi
fi

but ..azure-pipelines/azure-pipelines-linux.yml contains:

jobs:
- job: linux
  pool:
    vmImage: ubuntu-16.04
  strategy:
    matrix:
      linux_:
        CONFIG: linux_
        UPLOAD_PACKAGES: 'True'
        DOCKER_IMAGE: condaforge/linux-anvil-comp7
    maxParallel: 8
  timeoutInMinutes: 360

which take precedence over the run_docker_build.sh.

I could edit manually azure-pipelines-linux.yml but it would be erased at each new regenerate. Any idea?

[jakirkham]

It’s because of this.

[isuruf]

You need to do something like,

docker_image:
  - condaforge/linux-anvil-cos7-x86_64
cuda_compiler_version:
  - None