Fix bugs about the exit code (#262)

-   **fix: Make `fail: false` effective even when `failIfEmpty: true`**

This commit also makes sure `outputs.exit_code` is “The exit code returned from Lychee”. `failIfEmpty` no longer changes it to `1`.

Relevant docs:
- [Setting exit codes for actions - GitHub Docs](https://docs.github.com/en/actions/sharing-automations/creating-actions/setting-exit-codes-for-actions)
- [exit - POSIX Programmer's Manual](https://manned.org/exit.1posix)

Relates to #86, #128, #145, #245, and #251.

-   **fix: Update `env.exit_code` to `outputs.exit_code`**

The previous expression always gives `false`.
Both `env.exit_code` and `env.lychee_exit_code` are `null`, probably since the docker→composite refactor #128. When GitHub evaluates the expression, it finds the types do not match, and coerces them to number, namely, `null` → `0`.

See [Evaluate expressions in workflows and actions - GitHub Docs](https://docs.github.com/en/actions/writing-workflows/choosing-what-your-workflow-does/evaluate-expressions-in-workflows-and-actions#operators).

Relates to #253.



Co-authored-by: Sebastiaan Speck <[email protected]>

.github/workflows/links.yml

@@ -15,6 +15,7 @@ jobs:
       - uses: actions/checkout@v4
 
       - name: Link Checker
+        id: lychee
         uses: ./ # Uses an action in the root directory
         with:
           args: --user-agent "Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:59.0) Gecko/20100101 Firefox/59.0" --verbose --exclude spinroot.com --no-progress './**/*.md' './**/*.html' './**/*.rst'
@@ -23,7 +24,7 @@ jobs:
           GITHUB_TOKEN: ${{secrets.GITHUB_TOKEN}}
 
       - name: Create Issue From File
-        if: env.exit_code != 0
+        if: ${{ steps.lychee.outputs.exit_code }} != 0
         uses: peter-evans/create-issue-from-file@v5
         with:
           title: Link Checker Report

.github/workflows/test.yml

@@ -125,6 +125,28 @@ jobs:
           echo "Output set in args and action input should have failed."
           exit 1
 
+      - name: test fail - a lychee error should fail the pipeline
+        id: fail_test
+        uses: ./
+        with:
+          args: --verbose --no-progress foo.bar
+          debug: true
+        continue-on-error: true
+
+      # Explicitly check the exit code of the previous step
+      # as it's expected to fail
+      - name: Check fail
+        if: steps.fail_test.outcome != 'failure'
+        run: |
+          echo "Fail should have failed because the URL is invalid."
+          exit 1
+
+      - name: test disable fail - it's okay if lychee throws an error
+        uses: ./
+        with:
+          args: --no-progress foo.bar
+          fail: false
+
       - name: test failIfEmpty - no links in input should fail the pipeline
         id: fail_if_empty_test
         uses: ./
@@ -147,6 +169,40 @@ jobs:
           args: --no-progress fixtures/empty.md
           failIfEmpty: false
 
+      - name: test disable failIfEmpty - a lychee error should still fail the pipeline
+        id: fail_but_not_failIfEmpty_test
+        uses: ./
+        with:
+          args: --verbose --no-progress foo.bar
+          failIfEmpty: false
+          debug: true
+        continue-on-error: true
+
+      # Explicitly check the exit code of the previous step
+      # as it's expected to fail
+      - name: Check fail when failIfEmpty is disabled
+        if: steps.fail_but_not_failIfEmpty_test.outcome != 'failure'
+        run: |
+          echo "Fail should have failed because the URL is invalid, even though failIfEmpty is disabled."
+          exit 1
+
+      - name: test disable fail - no links in input should still fail the pipeline
+        id: failIfEmpty_but_not_fail_test
+        uses: ./
+        with:
+          args: --verbose --no-progress fixtures/empty.md
+          fail: false
+          debug: true
+        continue-on-error: true
+
+      # Explicitly check the exit code of the previous step
+      # as it's expected to fail
+      - name: Check failIfEmpty when fail is disabled
+        if: steps.failIfEmpty_but_not_fail_test.outcome != 'failure'
+        run: |
+          echo "FailIfEmpty should have failed because no links were found, even though fail is disabled."
+          exit 1
+
       - name: Install jq
         run: sudo apt-get install jq
 

README.md

@@ -37,7 +37,7 @@ jobs:
         uses: lycheeverse/lychee-action@v2
 
       - name: Create Issue From File
-        if: env.exit_code != 0
+        if: ${{ steps.lychee.outputs.exit_code }} != 0
         uses: peter-evans/create-issue-from-file@v5
         with:
           title: Link Checker Report

action.yml

@@ -106,20 +106,7 @@ runs:
 
     - name: Run Lychee
       id: run-lychee
-      run: |
-        # This step runs lychee and captures its exit code.
-        # We use 'set +e' to prevent the script from exiting immediately if lychee fails.
-        # This allows us to capture the exit code and pass it both to GitHub Actions (via GITHUB_OUTPUT)
-        # and to the shell (via the final 'exit $EXIT_CODE').
-        # This ensures that:
-        # 1. The step fails if lychee fails
-        # 2. The exit code is available as an output for subsequent steps
-        # 3. The exit code is properly propagated to the workflow
-        set +e
-        ${{ github.action_path }}/entrypoint.sh
-        EXIT_CODE=$?
-        echo "exit_code=$EXIT_CODE" >> $GITHUB_OUTPUT
-        exit $EXIT_CODE
+      run: ${{ github.action_path }}/entrypoint.sh
       env:
         # https://github.com/actions/runner/issues/665
         INPUT_TOKEN: ${{ inputs.TOKEN }}

entrypoint.sh

@@ -1,6 +1,13 @@
 #!/bin/bash -l
 set -uo pipefail
 
+# We use ‘set +e’ to prevent the script from exiting immediately if lychee fails.
+# This ensures that:
+# 1. Lychee exit code can be captured and passed to subsequent steps via `$GITHUB_OUTPUT`.
+# 2. This step’s outcome (success/failure) can be controlled according to inputs
+#    by manually calling the ‘exit’ command.
+set +e
+
 # Enable optional debug output
 if [ "${INPUT_DEBUG}" = true ]; then
   echo "Debug output enabled"
@@ -29,19 +36,17 @@ fi
 
 # Execute lychee
 eval lychee ${FORMAT} --output ${LYCHEE_TMP} ${ARGS} 
-exit_code=$?
+LYCHEE_EXIT_CODE=$?
 
-# Overwrite the exit code in case no links were found
-# and `failIfEmpty` is set to `true` (and it is by default)
+# If no links were found and `failIfEmpty` is set to `true` (and it is by default),
+# fail with an error later, but leave lychee exit code untouched.
+should_fail_because_empty=false
 if [ "${INPUT_FAILIFEMPTY}" = "true" ]; then
-    # Explicitly set INPUT_FAIL to true to ensure the script fails
-    # if no links are found
-    INPUT_FAIL=true
     # This is a somewhat crude way to check the Markdown output of lychee
     if grep -E 'Total\s+\|\s+0' "${LYCHEE_TMP}"; then
         echo "No links were found. This usually indicates a configuration error." >> "${LYCHEE_TMP}"
         echo "If this was expected, set 'failIfEmpty: false' in the args." >> "${LYCHEE_TMP}"
-        exit_code=1
+        should_fail_because_empty=true
     fi
 fi
 
@@ -67,12 +72,16 @@ if [ "${INPUT_FORMAT}" == "markdown" ]; then
   fi
 fi
 
-# Pass lychee exit code to next step
-echo "exit_code=$exit_code" >> $GITHUB_OUTPUT
+# Pass lychee exit code to subsequent steps
+echo "exit_code=$LYCHEE_EXIT_CODE" >> "$GITHUB_OUTPUT"
 
-# If `fail` is set to `true` (and it is by default), propagate the real exit
-# value to the workflow runner. This will cause the pipeline to fail on 
-# `exit != # 0`.
-if [ "$INPUT_FAIL" = true ] ; then
-    exit ${exit_code}
+# Determine the outcome of this step
+# Exiting with a nonzero value will fail the pipeline, but the specific value
+# does not matter. (GitHub does not share it with subsequent steps for composite actions.)
+if [ "$should_fail_because_empty" = true ] ; then
+  # If we decided previously to fail because no links were found, fail
+  exit 1
+elif [ "$INPUT_FAIL" = true ] ; then
+  # If `fail` is set to `true` (and it is by default), propagate lychee exit code
+  exit ${LYCHEE_EXIT_CODE}
 fi