Docker Build fails but Gitlab-CI reports succeeds(False positive) - docker

Question

I have a Gitlab pipeline that builds my Docker image from a Dockerfile, but when the "docker build" command fails, the pipeline still reports a success.
build:
stage: build
script:
- docker build --no-cache -t $CI_REGISTRY/dockerfile:$CONTAINER_LABEL .
I've added an error to my Dockerfile so I can provoke an error:
FROM ubuntu:20.04
RUN not_a_real_command_that_should_fail
The pipeline runs:
Running with gitlab-runner 14.0.1 (c1edb478)
on ******* Cy33WtLD
Preparing the "shell" executor
00:00
Using Shell executor...
Preparing environment
00:01
Running on **********...
Getting source from Git repository
00:04
Fetching changes with git depth set to 50...
Reinitialized existing Git repository in C:/gitlab-runner/builds/Cy33WtLD/0/dockerfile/.git/
Checking out 999a815d as fix_pipeline_status...
git-lfs/2.13.2 (GitHub; windows amd64; go 1.14.13; git fc664697)
Skipping Git submodules setup
Executing "step_script" stage of the job script
00:03
$ docker build --no-cache -t $CI_REGISTRY/dockerfile:$CONTAINER_LABEL .
#1 [internal] load build definition from Dockerfile
#1 sha256:c544637cbaca3e93c2a8a8c00efd4f81ee45b1abd410d971af12de8dae21e8ea
#1 transferring dockerfile: 3.04kB done
#1 DONE 0.0s
#2 [internal] load .dockerignore
#2 sha256:ab745a167b371ba5e9380063cb278a7792a5838550b89f02f35d7f6a583fb548
#2 transferring context: 2B done
#2 DONE 0.0s
#3 [internal] load metadata for docker.io/library/ubuntu:20.04
#3 sha256:8e67b796a66f85f06793e026943184e32d365c77929e94d2ac98b34a1e1cb30e
#3 DONE 0.6s
#4 [ 1/17] FROM docker.io/library/ubuntu:20.04#sha256:9d6a8699fb5c9c39cf08a0871bd6219f0400981c570894cd8cbea30d3424a31f
#4 sha256:c8b7f784dc481f981cf0bc39c4d4e60a54a355d96ca108a13ffffa3bfa047067
#4 CACHED
#20 [internal] load build context
#20 sha256:d12ef8e847404a2cc9437d8099f4b73f215c48eb92002e759a5f264989ae3ace
#20 transferring context: 92B 0.0s done
#20 DONE 0.0s
#5 [ 2/17] RUN not_a_real_command_that_should_fail
#5 sha256:724c85340f260555ab116f9064ba3c7a2c16fe0af059ef5226df31545b30ddb6
#5 0.485 /bin/sh: 1: not_a_real_command_that_should_fail: not found
#5 ERROR: executor failed running [/bin/sh -c not_a_real_command_that_should_fail]: exit code: 127
------
> [ 2/17] RUN not_a_real_command_that_should_fail:
------
executor failed running [/bin/sh -c not_a_real_command_that_should_fail]: exit code: 127
Cleaning up file based variables
00:01
Job succeeded
The exit code is 127, I've added a -after_stage step to print out the return code "echo $?" and get "true" result back. Seems like this should be sufficient for the command to trigger a failure in the pipeline.
The gitlab-runner is a shell executor on a Windows-machine, if that matters.
Any suggestions?

Answer 1

1 - Gitlab issue
There is a bug for gitlab runner versions prior to 13.1.1 :
https://gitlab.com/gitlab-org/gitlab-runner/-/issues/26347
Where job is always successful when FF_USE_LEGACY_KUBERNETES_EXECUTION_STRATEGY is set to False
Solution seems to be updating.
2 - Script issue
Try to force exit 1 on error with
build:
stage: build
script:
- docker build --no-cache -t $CI_REGISTRY/dockerfile:$CONTAINER_LABEL . || exit 1
Edit 1
By default, Powershell continues its execution.
You can set those variables :
job:
stage: build
variables:
ErrorActionPreference: stop
script:
- docker build --no-cache -t $CI_REGISTRY/dockerfile:$CONTAINER_LABEL .
Please see this issue on github for more information :
https://gitlab.com/gitlab-org/gitlab-runner/-/issues/4683

Answer 2

After moving the pipeline to a gitlab-runner on a Ubuntu machine the pipeline fails as expected. Assuming the Windows-Powershell exit-codes does not trigger correctly