Github Actions: check steps status
You can get status using setps.STEPNAME.outcome property, associated to a check of success() or failure()
name: CI
on: [pull_request]
jobs:
myjob:
runs-on: ubuntu-latest
steps:
- name: Step 1
id: hello
run: <any>
continue-on-error: true
- name: Step 2
id: world
run: <any>
continue-on-error: true
- name: Check on failures
if: (${{ success() }} || ${{ failure() }}) && (${{ steps.hello.outcome }} == 'failure' || ${{ steps.world.outcome }} == 'failure')
run: exit 1
Update: The steps
context now contains detail about the execution of each step by default. Using the outcome
property of each step we can check the result of its execution.
name: CI
on: [pull_request]
jobs:
myjob:
runs-on: ubuntu-latest
steps:
- name: Step 1
id: hello
run: <any>
continue-on-error: true
- name: Step 2
id: world
run: <any>
continue-on-error: true
- name: Check on failures
if: steps.hello.outcome != 'success' || steps.world.outcome != 'success'
run: exit 1
Original answer
Looking at the documentation for the steps context, it doesn't look like it contains any information about the step other than outputs
. These must be explicitly defined by steps. That is why the steps context is empty {}
.
https://help.github.com/en/articles/contexts-and-expression-syntax-for-github-actions#steps-context
Unfortunately, as far as I can tell, there is no default status for a step that can be accessed. The solution involves manually defining a status output variable from each step.
name: CI
on: [pull_request]
jobs:
myjob:
runs-on: ubuntu-latest
steps:
- name: Step 1
id: hello
run: echo ::set-output name=status::failure
continue-on-error: true
- name: Step 2
id: world
run: echo ::set-output name=status::success
continue-on-error: true
- name: Dump steps context
env:
STEPS_CONTEXT: ${{ toJson(steps) }}
run: echo "$STEPS_CONTEXT"
- name: Check on failures
if: steps.hello.outputs.status == 'failure' || steps.world.outputs.status == 'failure'
run: exit 1
This creates the following context output and the job fails.
{
"hello": {
"outputs": {
"status": "failure"
}
},
"world": {
"outputs": {
"status": "success"
}
}
}
https://help.github.com/en/articles/metadata-syntax-for-github-actions#outputs https://help.github.com/en/articles/development-tools-for-github-actions#set-an-output-parameter-set-output
jobs:
build:
name: Build
env:
DOCKER_PASS: ${{ secrets.DOCKER_PASS }}
runs-on: ubuntu-latest
steps:
- name: script
id: test1
continue-on-error: true
run: |
ls -l sadasd
- name: script
id: sync
continue-on-error: true
run: |
ls -l
- name: Dump steps context
env:
STEPS_CONTEXT: ${{ toJson(steps) }}
run: echo "$STEPS_CONTEXT"
the output
Run echo "$STEPS_CONTEXT"
{
"test1": {
"outputs": {},
"outcome": "failure",
"conclusion": "success"
},
"sync": {
"outputs": {},
"outcome": "success",
"conclusion": "success"
}
}
so could use step:
- name: sync run
id: sync
continue-on-error: true
run: |
.....
- name: after_success
run: |
...
if: steps.sync.outcome == 'success'
- name: after_failure
run: |
...
if: steps.sync.outcome != 'success'
I'm using somewhat similar to @peterevans suggested earlier but leveraging the shell exit for a command and set +e
flag:
name: CI
on: [pull_request]
jobs:
myjob:
runs-on: ubuntu-latest
steps:
- name: Step 1
id: hello
run: |
set +e
./my-script.sh
echo ::set-output name=exit_status::failure
- name: Step 2
id: world
run:
set +e
python3 ./my-script.py
echo ::set-output name=exit_status::$?
- name: Check on failures
if: steps.hello.outputs.exit_status != 0 | steps.world.outputs.exit_status != 0
run: exit 1