Fix: Preserve exit codes from shell commands and workflows (#1660)

* Changes auto-committed by Conductor

* Changes auto-committed by Conductor

* Changes auto-committed by Conductor

* updates

* updates

* Changes auto-committed by Conductor

* updates

* updates

* updates

* Fix duplicate exit code error messages using error wrapping pattern…

* [autofix.ci] apply automated fixes

* Fix duplicate exit code error messages by moving OsExit to errors package

- Move OsExit variable from pkg/utils to errors package to avoid import cycle
- Update all references to use errUtils.OsExit instead of utils.OsExit
- This fixes the issue where tests were failing because Exit() called os.Exit() directly
- The mockable OsExit is now in the errors package where Exit() function resides
- Resolves duplicate error message output in test failures

* Fix duplicate error message in shell runner exit code handling

- Remove errors.Join() that was combining ExitCodeError with original error
- Return only ExitCodeError to avoid duplicate message like 'subcommand exited with code 1 exit status 1'
- Update golden snapshot to reflect correct error message format
- Error message is now just 'subcommand exited with code 1' without duplication

* Remove temporary development files

- Remove check_fix_all_worktrees.sh (temporary worktree debugging script)
- Remove fix_worktree_indexes.sh (temporary worktree repair script)
- Remove test-fixes.patch (temporary patch file for testing)

These were development artifacts that should not be in the repository.

* Fix markdown linting issues in auth list blog post

- Add 'text' language identifier to code block (fixes MD040)
- Replace bold emphasis with proper heading syntax (fixes MD036)
- Convert **Quick Overview**, **Detailed Tree View**, and **Automation Integration** to #### headings
- Maintain consistent heading hierarchy within the document

* Fix duplicate error printing in workflow execution

- Make IsKnownWorkflowError recognize ExitCodeError as a known/handled error
- This prevents error from being printed twice (once by workflow, once by main.go)
- Return original error from workflow step failure to preserve exit code
- Workflow context error is already printed before returning
- Maintains error chain for proper exit code handling without duplication

* Remove auth-list blog post (covered by separate PR)

This file was accidentally included in this branch and is covered by the
osterman/auth-list-cmd PR. Removing to keep this PR focused on exit code fixes.

* Fix workspace creation to detect ExitCodeError instead of ExitError

- Updated terraform workspace handling to check for ExitCodeError with code 1
- Previously checked for os/exec.ExitError but ExecuteShellCommand now returns ExitCodeError
- This fixes tests failing with 'Workspace doesn't exist' errors
- Workspace creation fallback now works correctly when workspace select fails
- Resolves test failures: terraform_output_function, env_function, terraform_apply_env, etc.

* Update circuit-breaker test snapshot to remove trailing whitespace

- Regenerated snapshot for TestCLICommands/atmos_circuit-breaker test
- Removed excessive trailing whitespace from error message lines
- Snapshot now matches actual output without trailing spaces
- Error message 'subcommand exited with code 1' is now clean

* updates

* updates

---------

Co-authored-by: aknysh <[email protected]>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: Andriy Knysh <[email protected]>

Author: 3 people

.goreleaser.yml

@@ -1,8 +1,29 @@
 # Visit https://goreleaser.com for documentation on how to customize this behavior.
 
-#before:
-#  hooks:
-#  - 'go mod tidy'
+before:
+  hooks:
+    # Free disk space on GitHub Actions runners before building to prevent "no space left on device" errors.
+    # Building 14 platform binaries (linux, darwin, windows, freebsd × multiple architectures) requires
+    # significant temporary disk space for the Go linker.
+    #
+    # The default ubuntu-latest runner (currently ubuntu-24.04 or ubuntu-22.04) does come preloaded with many development tools that consume disk space:
+    #
+    # Category                    Approx. Size    Notes
+    # System files                ~20 GB          Base Ubuntu OS + required dependencies
+    # Android SDK / NDK           ~12 GB          Installed under /usr/local/lib/android — large but unused for most workflows
+    # .NET SDK                    ~2–3 GB         Installed under /usr/share/dotnet
+    # Haskell (GHC + Cabal)       ~2 GB           Installed under /opt/ghc
+    # CodeQL                      ~6 GB           Installed under /opt/hostedtoolcache/CodeQL — used for GitHub's code scanning
+    # Other languages/tools       ~15–20 GB       Includes Node.js, Python, Ruby, Go, Java, etc.
+    # Free space at job start     ~14 GB          You typically get 13–15 GB of free disk on /
+    - |
+      sh -c 'if [ "$GITHUB_ACTIONS" = "true" ]; then
+        echo "Freeing disk space..."
+        df -h /
+        sudo rm -rf /usr/local/lib/android /usr/share/dotnet /opt/ghc 2>/dev/null || true
+        df -h /
+      fi'
+
 builds:
   - env:
       # goreleaser does not work with CGO, it could also complicate
@@ -23,7 +44,14 @@ builds:
     binary: atmos
     ldflags:
       # Set `atmos` version to the GitHub release tag using Go `ldflags`
+      # -s: strip symbol table
+      # -w: strip DWARF debug info
+      # Both reduce binary size and disk usage during linking.
       - '-s -w -X "github.com/cloudposse/atmos/pkg/version.Version={{.Version}}"'
+    flags:
+      # Remove file system paths from compiled executables for reproducible builds and slightly smaller binaries.
+      # This is standard practice for open-source and CI/CD builds (used by Go core team, HashiCorp, Docker, etc.).
+      - -trimpath
 
 archives:
   - format: binary

cmd/root.go

@@ -114,7 +114,7 @@ var RootCmd = &cobra.Command{
 				if versionErr != nil {
 					errUtils.CheckErrorPrintAndExit(versionErr, "", "")
 				}
-				utils.OsExit(0)
+				errUtils.OsExit(0)
 				return
 			}
 		}
@@ -699,7 +699,7 @@ func initCobraConfig() {
 			pager := pager.NewWithAtmosConfig(pagerEnabled)
 			if err := pager.Run("Atmos CLI Help", buf.String()); err != nil {
 				log.Error("Failed to run pager", "error", err)
-				utils.OsExit(1)
+				errUtils.OsExit(1)
 			}
 		} else {
 			fmt.Println()

cmd/root_test.go

@@ -10,9 +10,9 @@ import (
 	"github.com/spf13/cobra"
 	"github.com/stretchr/testify/assert"
 
+	errUtils "github.com/cloudposse/atmos/errors"
 	log "github.com/cloudposse/atmos/pkg/logger"
 	"github.com/cloudposse/atmos/pkg/schema"
-	"github.com/cloudposse/atmos/pkg/utils"
 )
 
 func TestNoColorLog(t *testing.T) {
@@ -392,9 +392,9 @@ func TestVersionFlagExecutionPath(t *testing.T) {
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			// Save original OsExit and restore after test.
-			originalOsExit := utils.OsExit
+			originalOsExit := errUtils.OsExit
 			t.Cleanup(func() {
-				utils.OsExit = originalOsExit
+				errUtils.OsExit = originalOsExit
 				tt.cleanup()
 			})
 
@@ -403,7 +403,7 @@ func TestVersionFlagExecutionPath(t *testing.T) {
 			type exitPanic struct {
 				code int
 			}
-			utils.OsExit = func(code int) {
+			errUtils.OsExit = func(code int) {
 				panic(exitPanic{code: code})
 			}
 
@@ -414,7 +414,7 @@ func TestVersionFlagExecutionPath(t *testing.T) {
 				// Execute should call version command and then exit with expected code.
 				// We expect it to panic with our exitPanic struct containing the exit code.
 				// This verifies that the --version flag handler is being executed and
-				// calls os.Exit via utils.OsExit.
+				// calls os.Exit via errUtils.OsExit.
 				assert.PanicsWithValue(t, exitPanic{code: tt.expectExit}, func() {
 					_ = Execute()
 				}, "Execute should exit with code %d", tt.expectExit)

cmd/validate_schema.go

@@ -3,11 +3,11 @@ package cmd
 import (
 	"errors"
 
-	log "github.com/cloudposse/atmos/pkg/logger"
 	"github.com/spf13/cobra"
 
+	errUtils "github.com/cloudposse/atmos/errors"
 	"github.com/cloudposse/atmos/internal/exec"
-	u "github.com/cloudposse/atmos/pkg/utils"
+	log "github.com/cloudposse/atmos/pkg/logger"
 )
 
 // ValidateSchemaCmd represents the 'atmos validate schema' command.
@@ -59,12 +59,12 @@ and are compliant with expected formats, reducing configuration drift and runtim
 
 		if key == "" && schema != "" {
 			log.Error("key not provided for the schema to be used")
-			u.OsExit(1)
+			errUtils.OsExit(1)
 		}
 
 		if err := exec.NewAtmosValidatorExecutor(&atmosConfig).ExecuteAtmosValidateSchemaCmd(key, schema); err != nil {
 			if errors.Is(err, exec.ErrInvalidYAML) {
-				u.OsExit(1)
+				errUtils.OsExit(1)
 			}
 			return err
 		}

cmd/workflow.go

@@ -1,6 +1,8 @@
 package cmd
 
 import (
+	"errors"
+
 	"github.com/spf13/cobra"
 
 	errUtils "github.com/cloudposse/atmos/errors"
@@ -41,6 +43,11 @@ var workflowCmd = &cobra.Command{
 			// Check if it's a known error that's already printed in ExecuteWorkflowCmd.
 			// If it is, we don't need to print it again, but we do need to exit with a non-zero exit code.
 			if e.IsKnownWorkflowError(err) {
+				// Check if the error wraps an ExitCodeError to preserve the actual exit code.
+				var exitCodeErr errUtils.ExitCodeError
+				if errors.As(err, &exitCodeErr) {
+					errUtils.Exit(exitCodeErr.Code)
+				}
 				errUtils.Exit(1)
 			}
 			return err

errors/error_funcs.go

@@ -5,14 +5,17 @@ import (
 	"os"
 	"os/exec"
 
-	log "github.com/cloudposse/atmos/pkg/logger"
 	"golang.org/x/text/cases"
 	"golang.org/x/text/language"
 
+	log "github.com/cloudposse/atmos/pkg/logger"
 	"github.com/cloudposse/atmos/pkg/schema"
 	"github.com/cloudposse/atmos/pkg/ui/markdown"
 )
 
+// OsExit is a variable for testing, so we can mock os.Exit.
+var OsExit = os.Exit
+
 // render is the global Markdown renderer instance initialized via InitializeMarkdown.
 var render *markdown.Renderer
 
@@ -59,6 +62,12 @@ func CheckErrorPrintAndExit(err error, title string, suggestion string) {
 
 	CheckErrorAndPrint(err, title, suggestion)
 
+	// Check for ExitCodeError (from ShellRunner preserving interp.ExitStatus)
+	var exitCodeErr ExitCodeError
+	if errors.As(err, &exitCodeErr) {
+		Exit(exitCodeErr.Code)
+	}
+
 	// Find the executed command's exit code from the error
 	var exitError *exec.ExitError
 	if errors.As(err, &exitError) {
@@ -73,5 +82,5 @@ func CheckErrorPrintAndExit(err error, title string, suggestion string) {
 
 // Exit exits the program with the specified exit code.
 func Exit(exitCode int) {
-	os.Exit(exitCode)
+	OsExit(exitCode)
 }

errors/error_funcs_test.go

@@ -42,6 +42,102 @@ func Test_CheckErrorPrintAndExit(t *testing.T) {
 	}
 }
 
+func TestPrintErrorMarkdownAndExit(t *testing.T) {
+	if os.Getenv("TEST_EXIT") == "1" {
+		er := errors.New("critical failure")
+		CheckErrorPrintAndExit(er, "Fatal Error", "Check logs.")
+		return
+	}
+	execPath, err := exec.LookPath(os.Args[0])
+	assert.Nil(t, err)
+	cmd := exec.Command(execPath, "-test.run=TestPrintErrorMarkdownAndExit")
+	cmd.Env = append(os.Environ(), "TEST_EXIT=1")
+	err = cmd.Run()
+	var exitError *exec.ExitError
+	if errors.As(err, &exitError) {
+		assert.Equal(t, 1, exitError.ExitCode())
+	} else {
+		assert.Fail(t, "Expected an exit error with code 1")
+	}
+}
+
+func TestPrintInvalidUsageErrorAndExit(t *testing.T) {
+	if os.Getenv("TEST_EXIT") == "1" {
+		er := errors.New("invalid command")
+		CheckErrorPrintAndExit(er, "", "Use --help for usage information.")
+		return
+	}
+	execPath, err := exec.LookPath(os.Args[0])
+	assert.Nil(t, err)
+	cmd := exec.Command(execPath, "-test.run=TestPrintInvalidUsageErrorAndExit")
+	cmd.Env = append(os.Environ(), "TEST_EXIT=1")
+	err = cmd.Run()
+	var exitError *exec.ExitError
+	if errors.As(err, &exitError) {
+		assert.Equal(t, 1, exitError.ExitCode())
+	} else {
+		assert.Fail(t, "Expected an exit error with code 1")
+	}
+}
+
+func TestCheckErrorPrintAndExit_ExitCodeError(t *testing.T) {
+	if os.Getenv("TEST_EXIT_CODE_2") == "1" {
+		err := ExitCodeError{Code: 2}
+		CheckErrorPrintAndExit(err, "Exit Code Error", "")
+		return
+	}
+	if os.Getenv("TEST_EXIT_CODE_42") == "1" {
+		err := ExitCodeError{Code: 42}
+		CheckErrorPrintAndExit(err, "Exit Code Error", "")
+		return
+	}
+
+	// Test exit code 2
+	execPath, err := exec.LookPath(os.Args[0])
+	assert.Nil(t, err)
+	cmd := exec.Command(execPath, "-test.run=TestCheckErrorPrintAndExit_ExitCodeError")
+	cmd.Env = append(os.Environ(), "TEST_EXIT_CODE_2=1")
+	err = cmd.Run()
+	var exitError *exec.ExitError
+	if errors.As(err, &exitError) {
+		assert.Equal(t, 2, exitError.ExitCode(), "Should exit with code 2")
+	} else {
+		assert.Fail(t, "Expected an exit error with code 2")
+	}
+
+	// Test exit code 42
+	cmd = exec.Command(execPath, "-test.run=TestCheckErrorPrintAndExit_ExitCodeError")
+	cmd.Env = append(os.Environ(), "TEST_EXIT_CODE_42=1")
+	err = cmd.Run()
+	if errors.As(err, &exitError) {
+		assert.Equal(t, 42, exitError.ExitCode(), "Should exit with code 42")
+	} else {
+		assert.Fail(t, "Expected an exit error with code 42")
+	}
+}
+
+func TestCheckErrorPrintAndExit_ExecExitError(t *testing.T) {
+	if os.Getenv("TEST_EXEC_EXIT") == "1" {
+		// Create an exec.ExitError
+		cmd := exec.Command("sh", "-c", "exit 3")
+		err := cmd.Run()
+		CheckErrorPrintAndExit(err, "Exec Error", "")
+		return
+	}
+
+	execPath, err := exec.LookPath(os.Args[0])
+	assert.Nil(t, err)
+	cmd := exec.Command(execPath, "-test.run=TestCheckErrorPrintAndExit_ExecExitError")
+	cmd.Env = append(os.Environ(), "TEST_EXEC_EXIT=1")
+	err = cmd.Run()
+	var exitError *exec.ExitError
+	if errors.As(err, &exitError) {
+		assert.Equal(t, 3, exitError.ExitCode(), "Should exit with code 3 from exec.ExitError")
+	} else {
+		assert.Fail(t, "Expected an exit error with code 3")
+	}
+}
+
 func TestCheckErrorAndPrint(t *testing.T) {
 	// Save original logger
 	originalLogger := log.Default()
@@ -96,42 +192,47 @@ func TestCheckErrorAndPrint(t *testing.T) {
 		// Should log error directly
 		assert.Contains(t, logBuf.String(), "error with nil render")
 	})
-}
 
-func TestPrintErrorMarkdownAndExit(t *testing.T) {
-	if os.Getenv("TEST_EXIT") == "1" {
-		er := errors.New("critical failure")
-		CheckErrorPrintAndExit(er, "Fatal Error", "Check logs.")
-		return
-	}
-	execPath, err := exec.LookPath(os.Args[0])
-	assert.Nil(t, err)
-	cmd := exec.Command(execPath, "-test.run=TestPrintErrorMarkdownAndExit")
-	cmd.Env = append(os.Environ(), "TEST_EXIT=1")
-	err = cmd.Run()
-	var exitError *exec.ExitError
-	if errors.As(err, &exitError) {
-		assert.Equal(t, 1, exitError.ExitCode())
-	} else {
-		assert.Fail(t, "Expected an exit error with code 1")
-	}
+	// Test with empty title (defaults to "Error")
+	t.Run("empty title", func(t *testing.T) {
+		render, _ = markdown.NewTerminalMarkdownRenderer(schema.AtmosConfiguration{})
+
+		r, w, _ := os.Pipe()
+		os.Stderr = w
+
+		testErr := errors.New("test error with empty title")
+		CheckErrorAndPrint(testErr, "", "")
+
+		w.Close()
+		os.Stderr = oldStderr
+
+		var output bytes.Buffer
+		io.Copy(&output, r)
+
+		// Should default to "Error" as title
+		assert.Contains(t, output.String(), "Error")
+	})
 }
 
-func TestPrintInvalidUsageErrorAndExit(t *testing.T) {
-	if os.Getenv("TEST_EXIT") == "1" {
-		er := errors.New("invalid command")
-		CheckErrorPrintAndExit(er, "", "Use --help for usage information.")
-		return
-	}
-	execPath, err := exec.LookPath(os.Args[0])
-	assert.Nil(t, err)
-	cmd := exec.Command(execPath, "-test.run=TestPrintErrorMarkdownAndExit")
-	cmd.Env = append(os.Environ(), "TEST_EXIT=1")
-	err = cmd.Run()
-	var exitError *exec.ExitError
-	if errors.As(err, &exitError) {
-		assert.Equal(t, 1, exitError.ExitCode())
-	} else {
-		assert.Fail(t, "Expected an exit error with code 1")
-	}
+func TestInitializeMarkdown(t *testing.T) {
+	// Save original logger
+	originalLogger := log.Default()
+	defer log.SetDefault(originalLogger)
+
+	// Create test logger to capture log output
+	var logBuf bytes.Buffer
+	testLogger := log.New()
+	testLogger.SetOutput(&logBuf)
+	testLogger.SetLevel(log.TraceLevel)
+	log.SetDefault(testLogger)
+
+	// Test with valid configuration
+	t.Run("valid configuration", func(t *testing.T) {
+		logBuf.Reset()
+		atmosConfig := schema.AtmosConfiguration{}
+		InitializeMarkdown(atmosConfig)
+
+		// Should initialize without error
+		assert.NotContains(t, logBuf.String(), "failed to initialize Markdown renderer")
+	})
 }