Preserve image replay compatibility across CLI image limits

.github/workflows/python-sdk-tests.yml

@@ -99,6 +99,16 @@ jobs:
         # running independent modules concurrently in isolated workers.
         run: uv run pytest -v -s -n 2 --dist=loadfile
 
+      - name: Upload Python test diagnostics
+        if: failure()
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: python-test-diagnostics-${{ matrix.os }}-${{ matrix.python-version }}-${{ matrix.transport }}-${{ github.run_attempt }}
+          path: python/.pytest-diagnostics/
+          include-hidden-files: true
+          if-no-files-found: warn
+          retention-days: 7
+
   # JavaScript actions use a glibc-linked Node runtime, so Alpine runs through Docker.
   test-musl-arm64:
     name: "Python SDK Tests (Alpine ARM64, ${{ matrix.transport }})"
@@ -136,3 +146,13 @@ jobs:
             cd python
             uv sync --all-extras --dev
             uv run pytest -v -s -n 2 --dist=loadfile
+
+      - name: Upload Python test diagnostics
+        if: failure()
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: python-test-diagnostics-alpine-arm64-${{ matrix.transport }}-${{ github.run_attempt }}
+          path: python/.pytest-diagnostics/
+          include-hidden-files: true
+          if-no-files-found: warn
+          retention-days: 7

go/internal/e2e/auto_tier_e2e_test.go

@@ -62,6 +62,9 @@ func TestAutoTierE2E(t *testing.T) {
 	}
 
 	t.Run("should stage and reset auto tier preference", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		session := newAutoSession(t)
 		assertNoPending(t, session)
 
@@ -107,6 +110,9 @@ func TestAutoTierE2E(t *testing.T) {
 	})
 
 	t.Run("should preserve auto tier when set model omits it", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		session := newAutoSession(t)
 
 		if _, err := session.SetAutoTier(t.Context(), autoTier(copilot.AutoTierBalance)); err != nil {

go/internal/e2e/client_api_e2e_test.go

@@ -62,9 +62,9 @@ func TestClientAPIE2E(t *testing.T) {
 	})
 
 	t.Run("should get null last session id before any sessions exist", func(t *testing.T) {
-		// Use a fresh client with isolated COPILOT_HOME so other subtests don't pollute state.
-		freshCtx := testharness.NewTestContext(t)
-		freshClient := freshCtx.NewClient()
+		freshClient := ctx.NewClient(func(options *copilot.ClientOptions) {
+			options.BaseDirectory = t.TempDir()
+		})
 		t.Cleanup(func() { freshClient.ForceStop() })
 
 		if err := freshClient.Start(t.Context()); err != nil {

go/internal/e2e/client_options_e2e_test.go

@@ -19,6 +19,9 @@ import (
 // in package-level unit tests.
 func TestClientOptionsE2E(t *testing.T) {
 	t.Run("should listen on configured TCP port", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		port := getAvailableTCPPort(t)
 
@@ -45,6 +48,9 @@ func TestClientOptionsE2E(t *testing.T) {
 	})
 
 	t.Run("should use client cwd for default workingdirectory", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
 
@@ -85,6 +91,9 @@ func TestClientOptionsE2E(t *testing.T) {
 	})
 
 	t.Run("should propagate process options to spawned cli", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		// Mirrors: Should_Propagate_Process_Options_To_Spawned_Cli
 		// Spawns a fake stdio CLI (a Node.js script) so we can assert that the
 		// SDK passes the right argv / env / cwd / RPC params through to the
@@ -233,6 +242,9 @@ func TestClientOptionsE2E(t *testing.T) {
 	})
 
 	t.Run("should send empty-mode custom agent locality defaults in initial requests", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		cliPath := filepath.Join(ctx.WorkDir, "fake-cli-empty-"+randomHex(t)+".js")
 		capturePath := filepath.Join(ctx.WorkDir, "fake-cli-empty-capture-"+randomHex(t)+".json")
@@ -301,6 +313,9 @@ func TestClientOptionsE2E(t *testing.T) {
 	})
 
 	t.Run("should forward advanced session creation options to the CLI", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		cliPath := filepath.Join(ctx.WorkDir, "fake-cli-"+randomHex(t)+".js")
 		capturePath := filepath.Join(ctx.WorkDir, "fake-cli-capture-"+randomHex(t)+".json")
@@ -449,6 +464,9 @@ func TestClientOptionsE2E(t *testing.T) {
 	})
 
 	t.Run("should forward singular provider configuration on session creation", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		cliPath := filepath.Join(ctx.WorkDir, "fake-cli-"+randomHex(t)+".js")
 		capturePath := filepath.Join(ctx.WorkDir, "fake-cli-capture-"+randomHex(t)+".json")
@@ -510,6 +528,9 @@ func TestClientOptionsE2E(t *testing.T) {
 	})
 
 	t.Run("should forward advanced session resume options to the CLI", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		cliPath := filepath.Join(ctx.WorkDir, "fake-cli-"+randomHex(t)+".js")
 		capturePath := filepath.Join(ctx.WorkDir, "fake-cli-capture-"+randomHex(t)+".json")

go/internal/e2e/main_test.go

@@ -0,0 +1,12 @@
+package e2e
+
+import (
+	"os"
+	"testing"
+
+	"github.com/github/copilot-sdk/go/internal/e2e/testharness"
+)
+
+func TestMain(m *testing.M) {
+	os.Exit(testharness.RunWithInProcessIsolation(m))
+}

go/internal/e2e/pre_mcp_tool_call_hook_e2e_test.go

@@ -11,10 +11,6 @@ import (
 )
 
 func TestPreMCPToolCallHookE2E(t *testing.T) {
-	ctx := testharness.NewTestContext(t)
-	client := ctx.NewClient()
-	t.Cleanup(func() { client.ForceStop() })
-
 	testHarnessDir := testharness.RepoPath("test", "harness")
 	metaEchoServer := filepath.Join(testHarnessDir, "test-mcp-meta-echo-server.mjs")
 
@@ -30,7 +26,13 @@ func TestPreMCPToolCallHookE2E(t *testing.T) {
 	}
 
 	t.Run("should set meta via preMcpToolCall hook", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
+		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
+		client := ctx.NewClient()
+		t.Cleanup(func() { client.ForceStop() })
 
 		var (
 			mu     sync.Mutex
@@ -93,7 +95,13 @@ func TestPreMCPToolCallHookE2E(t *testing.T) {
 	})
 
 	t.Run("should replace meta via preMcpToolCall hook", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
+		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
+		client := ctx.NewClient()
+		t.Cleanup(func() { client.ForceStop() })
 
 		var (
 			mu     sync.Mutex
@@ -149,7 +157,13 @@ func TestPreMCPToolCallHookE2E(t *testing.T) {
 	})
 
 	t.Run("should remove meta via preMcpToolCall hook", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
+		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
+		client := ctx.NewClient()
+		t.Cleanup(func() { client.ForceStop() })
 
 		var (
 			mu     sync.Mutex

go/internal/e2e/rpc_mcp_config_e2e_test.go

@@ -12,6 +12,9 @@ import (
 // Tests server-scoped MCP configuration management via MCP.Config.* RPCs.
 func TestRPCMCPConfigE2E(t *testing.T) {
 	t.Run("should call server MCP config rpcs", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		client := ctx.NewClient()
 		t.Cleanup(func() { client.ForceStop() })
@@ -105,6 +108,9 @@ func TestRPCMCPConfigE2E(t *testing.T) {
 	})
 
 	t.Run("should round trip http MCP oauth config rpc", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		client := ctx.NewClient()
 		t.Cleanup(func() { client.ForceStop() })

go/internal/e2e/rpc_server_e2e_test.go

@@ -18,6 +18,9 @@ import (
 // Tests server-scoped (non-session) RPCs.
 func TestRPCServerE2E(t *testing.T) {
 	t.Run("should clear the managed settings cache", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
 		client := ctx.NewClient()
@@ -33,6 +36,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should call rpc ping with typed params and result", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
 		client := ctx.NewClient()
@@ -56,6 +62,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should call rpc models list with typed result", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
 		const token = "rpc-models-token"
@@ -89,6 +98,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should call rpc account get quota when authenticated", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
 		const token = "rpc-quota-token"
@@ -145,6 +157,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should call rpc tools list with typed result", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
 		client := ctx.NewClient()
@@ -169,6 +184,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should call rpc session fs set provider with typed result", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		client := ctx.NewClient()
 		t.Cleanup(func() { client.ForceStop() })
@@ -192,6 +210,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should add secret filter values", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		client := ctx.NewClient(func(opts *copilot.ClientOptions) {
 			opts.Env = append(opts.Env, "COPILOT_ENABLE_SECRET_FILTERING=true")
@@ -213,6 +234,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should return false for missing LLM response frames", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		client := ctx.NewClient()
 		t.Cleanup(func() { client.ForceStop() })
@@ -251,6 +275,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should list find and inspect persisted session state", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		token := "rpc-server-list-token-" + randomHex(t)
 		registerProxyUser(t, ctx, token, "rpc-user", nil)
@@ -344,6 +371,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should enrich basic session metadata", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		token := "rpc-server-enrich-token-" + randomHex(t)
 		registerProxyUser(t, ctx, token, "rpc-user", nil)
@@ -397,6 +427,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should close active session and release lock", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		token := "rpc-server-close-token-" + randomHex(t)
 		registerProxyUser(t, ctx, token, "rpc-user", nil)
@@ -434,6 +467,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should prune dry run and bulk delete persisted session", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		token := "rpc-server-delete-token-" + randomHex(t)
 		registerProxyUser(t, ctx, token, "rpc-user", nil)
@@ -500,6 +536,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should set additional plugins and reload deferred hooks", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		client := ctx.NewClient()
 		t.Cleanup(func() { client.ForceStop() })
@@ -546,6 +585,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should report implemented error when connecting unknown remote session", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureWithoutSnapshot(t)
 		client := ctx.NewClient()
@@ -568,6 +610,9 @@ func TestRPCServerE2E(t *testing.T) {
 	})
 
 	t.Run("should discover server mcp and skills", func(t *testing.T) {
+		if testharness.RunInIsolatedProcess(t) {
+			return
+		}
 		ctx := testharness.NewTestContext(t)
 		ctx.ConfigureForTest(t)
 		client := ctx.NewClient()