From f46fc98110cb5fbbf0b3e26e9d2a36336a42837e Mon Sep 17 00:00:00 2001
From: Mikhail Chusavitin <mchusavitin@mchusmbp.local>
Date: Tue, 7 Jul 2026 15:31:31 +0300
Subject: [PATCH] sat: add tests for informational discovery jobs and retry
 logic

Covers the sat.go fix from 2599d9c: a failing "dcgmi discovery -l"
preflight job must not flip the pack's overall status, and jobs marked
with retries should recover from a transient first-attempt failure.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
---
 audit/internal/platform/sat_test.go | 62 +++++++++++++++++++++++++++++
 1 file changed, 62 insertions(+)

diff --git a/audit/internal/platform/sat_test.go b/audit/internal/platform/sat_test.go
index 701b771..c6d18b0 100644
--- a/audit/internal/platform/sat_test.go
+++ b/audit/internal/platform/sat_test.go
@@ -622,3 +622,65 @@ func TestRunROCmSMIReportsMissingCommand(t *testing.T) {
 		t.Fatal("expected missing rocm-smi error")
 	}
 }
+
+// TestRunAcceptancePackCtxInformationalJobDoesNotFailOverallStatus guards
+// against a real production bug: a preflight/metadata job like "dcgmi
+// discovery -l" racing nv-hostengine startup must not flip the whole pack's
+// overall status to FAILED when the actual diagnostic that follows it passes.
+func TestRunAcceptancePackCtxInformationalJobDoesNotFailOverallStatus(t *testing.T) {
+	dir := t.TempDir()
+
+	runDir, err := runAcceptancePackCtx(context.Background(), dir, "gpu-nvidia-informational-test", []satJob{
+		{name: "01-preflight.log", cmd: []string{"sh", "-c", "echo boom >&2; exit 1"}, informational: true},
+		{name: "02-real-test.log", cmd: []string{"sh", "-c", "echo ok"}},
+	}, nil)
+	if err != nil {
+		t.Fatalf("runAcceptancePackCtx error: %v", err)
+	}
+
+	summary, err := os.ReadFile(filepath.Join(runDir, "summary.txt"))
+	if err != nil {
+		t.Fatalf("reading summary.txt: %v", err)
+	}
+	got := string(summary)
+	if !strings.Contains(got, "overall_status=OK") {
+		t.Fatalf("summary=%q want overall_status=OK", got)
+	}
+	if !strings.Contains(got, "job_informational_failed=1") {
+		t.Fatalf("summary=%q want job_informational_failed=1", got)
+	}
+	if !strings.Contains(got, "job_failed=0") {
+		t.Fatalf("summary=%q want job_failed=0", got)
+	}
+}
+
+// TestRunAcceptancePackCtxRetriesJob guards the retry path added for jobs
+// racing a service's startup (e.g. dcgmi discovery immediately after nv-hostengine
+// restarts): a job that fails once but succeeds on retry should end up OK.
+func TestRunAcceptancePackCtxRetriesJob(t *testing.T) {
+	dir := t.TempDir()
+	marker := filepath.Join(dir, "attempted")
+
+	runDir, err := runAcceptancePackCtx(context.Background(), dir, "gpu-nvidia-retry-test", []satJob{
+		{
+			name:    "01-flaky.log",
+			cmd:     []string{"sh", "-c", "if [ -f " + marker + " ]; then echo ok; else touch " + marker + "; exit 1; fi"},
+			retries: 2,
+		},
+	}, nil)
+	if err != nil {
+		t.Fatalf("runAcceptancePackCtx error: %v", err)
+	}
+
+	summary, err := os.ReadFile(filepath.Join(runDir, "summary.txt"))
+	if err != nil {
+		t.Fatalf("reading summary.txt: %v", err)
+	}
+	got := string(summary)
+	if !strings.Contains(got, "overall_status=OK") {
+		t.Fatalf("summary=%q want overall_status=OK", got)
+	}
+	if !strings.Contains(got, "flaky_status=OK") {
+		t.Fatalf("summary=%q want flaky_status=OK", got)
+	}
+}