Latest version of Remote Containers causes an error when building my devcontainer

[PaulHiemstra]

The latest version of remote containers no longer allows me to build a container that is externally defined in a dockerfile. The issue is fixed by going back to the previous version of the plugin.

• VSCode Version:1.72.0
• Local OS Version: Windows 11 21H2
• Remote OS Version: Ubuntu 20.04 (WSL2)
• Remote Extension/Connection Type: Container running in Docker desktop
• Logs: error.log

Steps to Reproduce:

1. Create new directory and drop the .devcontainer.json file in the root

{
  "name": "Tensorflow with Jupyter (GPU enabled)",
  "build": { "dockerfile": "docker/Dockerfile" },
  "forwardPorts": [8888],
  "postStartCommand": "nohup bash -c 'jupyter notebook --ip=0.0.0.0 --port=8888 --allow-root --NotebookApp.token=jupdevtoken &'",
  "runArgs": [
    "--gpus",
    "all"
  ],
  "extensions": [
    "ms-python.python",
    "ms-python.vscode-pylance",
    "ms-toolsai.jupyter"
  ]
}

2. Create a docker subdirectory, drop Dockerfile in that subdir

FROM tensorflow/tensorflow:latest-gpu-jupyter

RUN pip install pandas\
    sklearn\
    seaborn\

3. Start VS Code and restart the directory in the container
4. The following error occurs when building the container (full log in error.log

    error: failed to solve: rpc error: code = Unknown desc = failed to solve with fr
    ontend dockerfile.v0: failed to solve with frontend gateway.v0: rpc error: code    
    = Unknown desc = target stage dev_containers_target_stage could not be found

Does this issue occur when you try this locally?: Yes
Does this issue occur when you try this locally and all extensions are disabled?: Yes

[chrmarti]

What do you get for docker buildx ls? If the current builder (marked with *) is not the default one, try switching to it with docker buildx use default and retry.

[chrmarti]

Sorry, ignore the previous comment, I misread the error message.

Could you append C:\Users\PH0110~1\AppData\Local\Temp\devcontainercli\container-features\0.19.0-1665383336028\Dockerfile-with-features mentioned in the log?

[PaulHiemstra]

Here is the Dockerfile you requested:
Dockerfile-with-features.txt

# syntax=docker/dockerfile:1.4
ARG _DEV_CONTAINERS_BASE_IMAGE=placeholder
FROM tensorflow/tensorflow:latest-gpu-jupyter AS dev_container_auto_added_stage_label
WORKDIR "/tmp"

RUN pip install pandas\
    sklearn\
    seaborn\
    mlflow\






FROM $_DEV_CONTAINERS_BASE_IMAGE AS dev_containers_target_stage

USER root

COPY --from=dev_containers_feature_content_source . /tmp/build-features/







ARG _DEV_CONTAINERS_IMAGE_USER=root
USER $_DEV_CONTAINERS_IMAGE_USER

LABEL devcontainer.metadata="{\"postStartCommand\":\"nohup bash -c 'jupyter notebook --ip=0.0.0.0 --port=8888 --allow-root --NotebookApp.token=jupdevtoken &'\",\"customizations\":{\"vscode\":{\"extensions\":[\"ms-python.python\",\"ms-python.vscode-pylance\",\"ms-toolsai.jupyter\"]}},\"forwardPorts\":[8888,5000]}"

[chrmarti]

The dev_containers_target_stage stage is there. Not sure why --target dev_containers_target_stage would not work.

[PaulHiemstra]

The issue must be in the recently added code, the release of a month ago does not show this issue.

[PaulHiemstra]

Here is a remote containers logfile of the version that works.
no_error.log.log

What strikes me as odd is that there is no mention of Dockerfile-with-features, so this seems to be a new addition that does not work correctly.

[AlexisLessard]

We're having the same issue, using this dockerfile:

# [Choice] .NET version: 6.0, 5.0, 3.1, 6.0-bullseye, 5.0-bullseye, 3.1-bullseye, 6.0-focal, 5.0-focal, 3.1-focal
ARG VARIANT=6.0
FROM mcr.microsoft.com/vscode/devcontainers/dotnet:0-${VARIANT}

#Install oc and kubectl client
RUN wget https://mirror.openshift.com/pub/openshift-v4/x86_64/clients/ocp/stable/openshift-client-linux.tar.gz \
    && tar -zxvf openshift-client-linux.tar.gz --directory /tmp/ \
    && cp /tmp/oc /tmp/kubectl /usr/local/sbin \
    && chmod +x /usr/local/sbin/oc && chmod +x /usr/local/sbin/kubectl \
    && rm -f /tmp/oc && rm -f /tmp/kubectl &&  rm -f /tmp/README.md

#Install ReportGenerator
RUN dotnet tool install -g dotnet-reportgenerator-globaltool \
    && dotnet tool install dotnet-reportgenerator-globaltool --tool-path tools \
    && dotnet new tool-manifest --name ReportGenerator \
    && dotnet tool install dotnet-reportgenerator-globaltool \

VSCode Version:1.72.1
Local OS Version: Windows 10 10.0.19044
Remote OS Version: Ubuntu 20.04 (WSL2)
Remote Extension/Connection Type: Container running in Docker desktop via the wsl2 backend. Folder is located in the WSL2 workspace, not on the windows file system.

[tjdett]

I had a similar problem. The issue is this line:

RUN pip install pandas\
    sklearn\
    seaborn\
    mlflow\

Docker may have previously supported a final trailing slash in a build file, but it appears to no longer be supported. (I'm not sure when this happened, but I'm on Docker 20.10.21 and it's no longer supported.)

This trailing slash appears to interfere with the following FROM statement.

Just remove the slash and it should work:

RUN pip install pandas\
    sklearn\
    seaborn\
    mlflow

[chrmarti]

@PaulHiemstra Does it work when you remove the trailing \ from your Dockerfile?

[AlexisLessard]

@chrmarti Seemed to do it for me! I can build my devcontainer again!

Any idea on why that is? What changed between 1.70 and 1.72 that created this problem?

[walsha2]

Does it work when you remove the trailing \ from your Dockerfile?

@chrmarti Can confirm as well. Removing the trailing \ from one of my Dockerfile commands (which previously was not an issue) worked.

Here is the error caused by that trailing \, same as @PaulHiemstra:

[2022-11-04T20:19:50.061Z] ERROR: failed to solve: rpc error: code = Unknown desc = failed 
to solve with frontend dockerfile.v0: failed to create LLB definition: 
target stage dev_containers_target_stage could not be found

---

@tjdett good find. Running on Docker version 20.10.17, build 100c701, M1 Macbook Pro

[sanmai-NL]

What's this name dev_containers_target_stage about? I also see this referenced, but that stage doesn‘t exist in my Dockerfile.