Don't let a broken fixture take down the whole test binary

A failing setup step called Shell.fail, which panicked. Tests run as
parallel subtests, so that panic aborted the entire test binary: one bad
fixture cost us the results of all ~500 tests, and the failure was
reported as a stack trace rather than against the test that caused it.

Keep panicking to skip the remaining setup steps -- they would only
produce follow-on failures -- but recover in createFixture and return the
message as that test's error. All three clients already propagate an
error from a test, so they report it the way they report any other
failure.

pkg/integration/components/runner.go

@@ -1,6 +1,7 @@
 package components
 
 import (
+	"errors"
 	"fmt"
 	"os"
 	"os/exec"
@@ -159,9 +160,7 @@ func prepareTestDir(
 		return "", err
 	}
 
-	workingDir := createFixture(test, paths, rootDir)
-
-	return workingDir, nil
+	return createFixture(test, paths, rootDir)
 }
 
 func buildLazygit(testArgs RunTestArgs) error {
@@ -182,22 +181,41 @@ func buildLazygit(testArgs RunTestArgs) error {
 	return osCommand.Cmd.New(args).Run()
 }
 
+// A failing setup step panics with this so that the remaining steps, which
+// would only produce follow-on failures, are skipped.
+type fixtureFailure string
+
 // Sets up the fixture for test and returns the working directory to invoke
 // lazygit in.
-func createFixture(test *IntegrationTest, paths Paths, rootDir string) string {
+func createFixture(test *IntegrationTest, paths Paths, rootDir string) (workingDir string, err error) {
+	// Tests run as parallel subtests, and a panic escaping one of them takes
+	// down the whole test binary, discarding every other test's result along
+	// with it. Report a broken fixture as this test's error instead.
+	defer func() {
+		panicValue := recover()
+		if panicValue == nil {
+			return
+		}
+		failure, ok := panicValue.(fixtureFailure)
+		if !ok {
+			panic(panicValue)
+		}
+		err = errors.New(string(failure))
+	}()
+
 	env := NewTestEnvironment(rootDir)
 
 	env = append(env, fmt.Sprintf("%s=%s", PWD, paths.ActualRepo()))
 	shell := NewShell(
 		paths.ActualRepo(),
 		env,
-		func(errorMsg string) { panic(errorMsg) },
+		func(errorMsg string) { panic(fixtureFailure(errorMsg)) },
 	)
 	shell.Init()
 
 	test.SetupRepo(shell)
 
-	return shell.dir
+	return shell.dir, nil
 }
 
 func testPath(rootdir string) string {

pkg/integration/components/test_test.go

@@ -1,8 +1,11 @@
 package components
 
 import (
+	"os"
+	"path/filepath"
 	"testing"
 
+	lazycoreUtils "github.com/jesseduffield/lazycore/pkg/utils"
 	"github.com/jesseduffield/lazygit/pkg/commands/git_commands"
 	"github.com/jesseduffield/lazygit/pkg/commands/models"
 	"github.com/jesseduffield/lazygit/pkg/config"
@@ -158,6 +161,27 @@ func TestSuccess(t *testing.T) {
 	assert.Equal(t, "", driver.failureMessage)
 }
 
+func TestFailingFixture(t *testing.T) {
+	test := NewIntegrationTest(NewIntegrationTestArgs{
+		Description: unitTestDescription,
+		SetupRepo: func(shell *Shell) {
+			shell.RunCommand([]string{"git", "checkout", "no-such-branch"})
+			shell.CreateFile("reached.txt", "")
+		},
+		Run: func(t *TestDriver, keys config.KeybindingConfig) {},
+	})
+
+	paths := NewPaths(t.TempDir())
+	assert.NoError(t, os.MkdirAll(paths.ActualRepo(), 0o777))
+
+	workingDir, err := createFixture(test, paths, lazycoreUtils.GetLazyRootDirectory())
+
+	assert.ErrorContains(t, err, "git checkout no-such-branch")
+	assert.Empty(t, workingDir)
+	// the steps following the failing one are skipped
+	assert.NoFileExists(t, filepath.Join(paths.ActualRepo(), "reached.txt"))
+}
+
 func TestGitVersionRestriction(t *testing.T) {
 	scenarios := []struct {
 		testName          string