test(harness): check fixture exit codes; fix exec signal exit codes and a ctrl-c panic

CHANGELOG.md

@@ -33,6 +33,11 @@ Begin agent-assisted development era.
 - README: improved Contributing section for both agents and humans
 - LICENSE: updated copyright to include Equinix Metal and 2025 A. Tobey
 
+### Fixed
+
+- `exec` exits 128 + signal when the child is killed by a signal, not 255 (Unix)
+- `exec` no longer panics on Ctrl-C during the span export
+
 ## [0.5.0] - 2025-04-21
 
 Fork otel-cli to @tobert's personal GitHub.

data_for_test.go

@@ -583,14 +583,27 @@ var suites = []FixtureSuite{
 				IsLongTest:           true,
 			},
 			Expect: Results{
-				Config: otelcli.DefaultConfig(),
+				Config:   otelcli.DefaultConfig(),
+				ExitCode: 42,
 			},
-			CheckFuncs: []CheckFunc{
-				func(t *testing.T, f Fixture, r Results) {
-					if r.ExitCode != 42 {
-						t.Errorf("expected exit code 42 from child process, got %d", r.ExitCode)
-					}
+		},
+		{
+			Name: "exec ctrl-c during export does not panic",
+			Config: FixtureConfig{
+				CliArgs: []string{"exec",
+					"--endpoint", "{{endpoint}}",
+					"--timeout", "1s",
+					"--", "/bin/true",
 				},
+				StopServerBeforeExec: true,
+				KillAfter:            time.Millisecond * 200, // child is done, export is retrying
+				KillSignal:           syscall.SIGINT,
+				TestTimeoutMs:        3000,
+				IsLongTest:           true,
+			},
+			Expect: Results{
+				Config:        otelcli.DefaultConfig(),
+				CommandFailed: true, // ended by the signal, not a panic (exit 2)
 			},
 		},
 	},
@@ -1155,14 +1168,14 @@ var suites = []FixtureSuite{
 				CliOutputRe: regexp.MustCompile(`^\d{4}/\d{2}/\d{2} \d{2}:\d{2}:\d{2} `),
 				CliOutput:   "invalid protocol setting \"xxx\"\n",
 				Config:      otelcli.DefaultConfig().WithEndpoint("{{endpoint}}"),
+				ExitCode:    1, // --fail
 				Diagnostics: otelcli.Diagnostics{
 					IsRecording:       false,
 					NumArgs:           7,
 					DetectedLocalhost: true,
 					ParsedTimeoutMs:   1000,
 					Endpoint:          "*",
 					EndpointSource:    "*",
-					ExecExitCode:      1,
 				},
 				SpanCount: 0,
 			},
@@ -1403,7 +1416,7 @@ var suites = []FixtureSuite{
 			Expect: Results{
 				SpanCount: 1,
 				Config:    otelcli.DefaultConfig().WithEndpoint("{{endpoint}}"),
-				ExitCode:  2,
+				ExitCode:  130, // 128 + SIGINT, as shells report it
 				SpanData: map[string]string{
 					"status_code":        "2",
 					"status_description": "exec command failed: signal: interrupt",
@@ -1423,7 +1436,7 @@ var suites = []FixtureSuite{
 			Expect: Results{
 				SpanCount: 1,
 				Config:    otelcli.DefaultConfig().WithEndpoint("{{endpoint}}"),
-				ExitCode:  2,
+				ExitCode:  137, // 128 + SIGKILL
 				SpanData: map[string]string{
 					"status_code":        "2",
 					"status_description": "exec command failed: signal: killed",
@@ -1444,7 +1457,7 @@ var suites = []FixtureSuite{
 			Expect: Results{
 				SpanCount: 1,
 				Config:    otelcli.DefaultConfig().WithEndpoint("{{endpoint}}"),
-				ExitCode:  0,
+				ExitCode:  137, // --command-timeout kills sleep at 100ms
 			},
 		},
 	},

main_test.go

@@ -207,6 +207,12 @@ func checkProcess(t *testing.T, fixture Fixture, results Results) bool {
 		t.Errorf("[%s] command failed is %t but expected %t", fixture.Name, results.CommandFailed, fixture.Expect.CommandFailed)
 		return false
 	}
+	// a killed process has no exit code of its own; TimedOut alone is not enough
+	// to skip, because the server timer sets it even when otel-cli exits normally
+	if !results.CommandFailed && results.ExitCode != fixture.Expect.ExitCode {
+		t.Errorf("[%s] exit code is %d but expected %d", fixture.Name, results.ExitCode, fixture.Expect.ExitCode)
+		return false
+	}
 	return true
 }
 

otelcli/exec.go

@@ -8,6 +8,7 @@ import (
 	"os/signal"
 	"os/user"
 	"strings"
+	"syscall"
 	"time"
 
 	"github.com/spf13/cobra"
@@ -156,11 +157,18 @@ func doExec(cmd *cobra.Command, args []string) {
 	// capture the child's exit code before OTLP export so SoftFail can use it (#360)
 	if child.ProcessState != nil {
 		Diag.ExecExitCode = child.ProcessState.ExitCode()
+		// a child killed by a signal reports -1, which os.Exit turns into 255;
+		// exit 128 + signal instead, as shells do
+		if status, ok := child.ProcessState.Sys().(syscall.WaitStatus); ok && status.Signaled() {
+			Diag.ExecExitCode = 128 + int(status.Signal())
+		}
 	} else {
 		Diag.ExecExitCode = 127
 	}
 
 	cancelCtxDeadline()
+	// stop delivery before close, or a ctrl-c during export panics on a closed channel
+	signal.Stop(signals)
 	close(signals)
 	<-signalsDone