Fix issues with subtasks attempting completion along with commands

evals/packages/types/src/roo-code.ts

@@ -758,7 +758,6 @@ export const isGlobalStateKey = (key: string): key is Keys<GlobalState> =>
 export const clineAsks = [
 	"followup",
 	"command",
-	"command_output",
 	"completion_result",
 	"tool",
 	"api_req_failed",

src/core/Cline.ts

@@ -161,7 +161,7 @@ export class Cline extends EventEmitter<ClineEvents> {
 	private askResponse?: ClineAskResponse
 	private askResponseText?: string
 	private askResponseImages?: string[]
-	private lastMessageTs?: number
+	public lastMessageTs?: number
 
 	// Not private since it needs to be accessible by tools.
 	consecutiveMistakeCount: number = 0
@@ -440,7 +440,6 @@ export class Cline extends EventEmitter<ClineEvents> {
 					*/
 					askTs = lastMessage.ts
 					this.lastMessageTs = askTs
-					// lastMessage.ts = askTs
 					lastMessage.text = text
 					lastMessage.partial = false
 					lastMessage.progressStatus = progressStatus
@@ -466,12 +465,20 @@ export class Cline extends EventEmitter<ClineEvents> {
 			await this.addToClineMessages({ ts: askTs, type: "ask", ask: type, text })
 		}
 
+		console.log(`[ask / ${type} / ${askTs}] waiting for ask response`)
+
 		await pWaitFor(() => this.askResponse !== undefined || this.lastMessageTs !== askTs, { interval: 100 })
 
+		const isTsMismatch = this.lastMessageTs !== askTs
+
+		console.log(
+			`[ask / ${type} / ${askTs}] pWaitFor returned, askResponse = ${this.askResponse}, isTsMismatch = ${isTsMismatch}`,
+		)
+
 		if (this.lastMessageTs !== askTs) {
-			// Could happen if we send multiple asks in a row i.e. with
-			// command_output. It's important that when we know an ask could
-			// fail, it is handled gracefully.
+			// Could happen if we send multiple asks in a row.
+			// It's important that when we know an ask could fail it is handled
+			// gracefully.
 			throw new Error("Current ask promise was ignored")
 		}
 

src/core/tools/attemptCompletionTool.ts

@@ -108,7 +108,7 @@ export async function attemptCompletionTool(
 				}
 
 				// tell the provider to remove the current subtask and resume the previous task in the stack
-				await cline.providerRef.deref()?.finishSubTask(lastMessage?.text ?? "")
+				await cline.providerRef.deref()?.finishSubTask(result)
 				return
 			}
 

src/core/tools/executeCommandTool.ts

@@ -9,7 +9,7 @@ import { ToolUse, AskApproval, HandleError, PushToolResult, RemoveClosingTag, To
 import { formatResponse } from "../prompts/responses"
 import { unescapeHtmlEntities } from "../../utils/text-normalization"
 import { telemetryService } from "../../services/telemetry/TelemetryService"
-import { ExitCodeDetails, RooTerminalCallbacks, RooTerminalProcess } from "../../integrations/terminal/types"
+import { ExitCodeDetails, RooTerminalCallbacks } from "../../integrations/terminal/types"
 import { TerminalRegistry } from "../../integrations/terminal/TerminalRegistry"
 import { Terminal } from "../../integrations/terminal/Terminal"
 
@@ -48,14 +48,15 @@ export async function executeCommandTool(
 
 			cline.consecutiveMistakeCount = 0
 
-			const executionId = Date.now().toString()
 			command = unescapeHtmlEntities(command) // Unescape HTML entities.
-			const didApprove = await askApproval("command", command, { id: executionId })
+			const didApprove = await askApproval("command", command)
 
 			if (!didApprove) {
 				return
 			}
 
+			const executionId = cline.lastMessageTs?.toString() ?? Date.now().toString()
+
 			const clineProvider = await cline.providerRef.deref()
 			const clineProviderState = await clineProvider?.getState()
 			const { terminalOutputLineLimit = 500, terminalShellIntegrationDisabled = false } = clineProviderState ?? {}
@@ -140,7 +141,6 @@ export async function executeCommand(
 	}
 
 	let message: { text?: string; images?: string[] } | undefined
-	let runInBackground = false
 	let completed = false
 	let result: string = ""
 	let exitDetails: ExitCodeDetails | undefined
@@ -150,23 +150,9 @@ export async function executeCommand(
 	const clineProvider = await cline.providerRef.deref()
 
 	const callbacks: RooTerminalCallbacks = {
-		onLine: async (output: string, process: RooTerminalProcess) => {
+		onLine: async (output: string) => {
 			const status: CommandExecutionStatus = { executionId, status: "output", output }
 			clineProvider?.postMessageToWebview({ type: "commandExecutionStatus", text: JSON.stringify(status) })
-
-			if (runInBackground) {
-				return
-			}
-
-			try {
-				const { response, text, images } = await cline.ask("command_output", "")
-				runInBackground = true
-
-				if (response === "messageResponse") {
-					message = { text, images }
-					process.continue()
-				}
-			} catch (_error) {}
 		},
 		onCompleted: (output: string | undefined) => {
 			result = Terminal.compressTerminalOutput(output ?? "", terminalOutputLineLimit)
@@ -177,6 +163,9 @@ export async function executeCommand(
 			console.log(`[executeCommand] onShellExecutionStarted: ${pid}`)
 			const status: CommandExecutionStatus = { executionId, status: "started", pid, command }
 			clineProvider?.postMessageToWebview({ type: "commandExecutionStatus", text: JSON.stringify(status) })
+			// This `command_output` message tells the webview to render the
+			// appropriate primary and secondary buttons.
+			cline.say("command_output", "")
 		},
 		onShellExecutionComplete: (details: ExitCodeDetails) => {
 			const status: CommandExecutionStatus = { executionId, status: "exited", exitCode: details.exitCode }
@@ -216,8 +205,7 @@ export async function executeCommand(
 	// Wait for a short delay to ensure all messages are sent to the webview.
 	// This delay allows time for non-awaited promises to be created and
 	// for their associated messages to be sent to the webview, maintaining
-	// the correct order of messages (although the webview is smart about
-	// grouping command_output messages despite any gaps anyways).
+	// the correct order of messages.
 	await delay(50)
 
 	if (message) {

src/core/webview/webviewMessageHandler.ts

@@ -184,7 +184,9 @@ export const webviewMessageHandler = async (provider: ClineProvider, message: We
 			await provider.postStateToWebview()
 			break
 		case "askResponse":
-			provider.getCurrentCline()?.handleWebviewAskResponse(message.askResponse!, message.text, message.images)
+			const instance = provider.getCurrentCline()
+			console.log("askResponse ->", message, !!instance)
+			instance?.handleWebviewAskResponse(message.askResponse!, message.text, message.images)
 			break
 		case "terminalOperation":
 			if (message.terminalOperation) {

src/exports/roo-code.d.ts

@@ -305,7 +305,6 @@ type ClineMessage = {
 		| (
 				| "followup"
 				| "command"
-				| "command_output"
 				| "completion_result"
 				| "tool"
 				| "api_req_failed"
@@ -381,7 +380,6 @@ type RooCodeEvents = {
 					| (
 							| "followup"
 							| "command"
-							| "command_output"
 							| "completion_result"
 							| "tool"
 							| "api_req_failed"

src/exports/types.ts

@@ -310,7 +310,6 @@ type ClineMessage = {
 		| (
 				| "followup"
 				| "command"
-				| "command_output"
 				| "completion_result"
 				| "tool"
 				| "api_req_failed"
@@ -390,7 +389,6 @@ type RooCodeEvents = {
 					| (
 							| "followup"
 							| "command"
-							| "command_output"
 							| "completion_result"
 							| "tool"
 							| "api_req_failed"

src/schemas/index.ts

@@ -767,7 +767,6 @@ export const isGlobalStateKey = (key: string): key is Keys<GlobalState> =>
 export const clineAsks = [
 	"followup",
 	"command",
-	"command_output",
 	"completion_result",
 	"tool",
 	"api_req_failed",

src/shared/__tests__/combineCommandSequences.test.ts

@@ -21,20 +21,13 @@ const messages: ClineMessage[] = [
 	},
 	{ ts: 1745710930748, type: "ask", ask: "command", text: "ping www.google.com", partial: false },
 	{ ts: 1745710930894, type: "say", say: "command_output", text: "", images: undefined },
-	{ ts: 1745710930894, type: "ask", ask: "command_output", text: "" },
 	{
 		ts: 1745710930954,
 		type: "say",
 		say: "command_output",
 		text: "PING www.google.com (142.251.46.228): 56 data bytes\n",
 		images: undefined,
 	},
-	{
-		ts: 1745710930954,
-		type: "ask",
-		ask: "command_output",
-		text: "PING www.google.com (142.251.46.228): 56 data bytes\n",
-	},
 ]
 
 describe("combineCommandSequences", () => {

src/shared/combineCommandSequences.ts

@@ -39,7 +39,7 @@ export function combineCommandSequences(messages: ClineMessage[]): ClineMessage[
 					break // Stop if we encounter the next command.
 				}
 
-				if (ask === "command_output" || say === "command_output") {
+				if (say === "command_output") {
 					if (!previous) {
 						combinedText += `\n${COMMAND_OUTPUT_STRING}`
 					}
@@ -68,7 +68,7 @@ export function combineCommandSequences(messages: ClineMessage[]): ClineMessage[
 	// Second pass: remove command_outputs and replace original commands with
 	// combined ones.
 	return messages
-		.filter((msg) => !(msg.ask === "command_output" || msg.say === "command_output"))
+		.filter((msg) => msg.say !== "command_output")
 		.map((msg) => {
 			if (msg.type === "ask" && msg.ask === "command") {
 				return combinedCommands.find((cmd) => cmd.ts === msg.ts) || msg

webview-ui/src/components/chat/ChatRow.tsx

@@ -985,7 +985,7 @@ export const ChatRowContent = ({
 								{icon}
 								{title}
 							</div>
-							<CommandExecution executionId={message.progressStatus?.id} text={message.text} />
+							<CommandExecution executionId={message.ts.toString()} text={message.text} />
 						</>
 					)
 				case "use_mcp_server":

webview-ui/src/components/chat/ChatView.tsx

@@ -116,7 +116,7 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 	const [selectedImages, setSelectedImages] = useState<string[]>([])
 
 	// we need to hold on to the ask because useEffect > lastMessage will always let us know when an ask comes in and handle it, but by the time handleMessage is called, the last message might not be the ask anymore (it could be a say that followed)
-	const [clineAsk, setClineAsk] = useState<ClineAsk | undefined>(undefined)
+	const [clineAsk, setClineAsk] = useState<ClineAsk | "command_output" | undefined>(undefined)
 	const [enableButtons, setEnableButtons] = useState<boolean>(false)
 	const [primaryButtonText, setPrimaryButtonText] = useState<string | undefined>(undefined)
 	const [secondaryButtonText, setSecondaryButtonText] = useState<string | undefined>(undefined)
@@ -229,13 +229,6 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 							setPrimaryButtonText(t("chat:runCommand.title"))
 							setSecondaryButtonText(t("chat:reject.title"))
 							break
-						case "command_output":
-							setTextAreaDisabled(false)
-							setClineAsk("command_output")
-							setEnableButtons(true)
-							setPrimaryButtonText(t("chat:proceedWhileRunning.title"))
-							setSecondaryButtonText(t("chat:killCommand.title"))
-							break
 						case "use_mcp_server":
 							if (!isAutoApproved(lastMessage) && !isPartial) {
 								playSound("notification")
@@ -289,8 +282,8 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 							setTextAreaDisabled(true)
 							break
 						case "api_req_started":
-							if (secondLastMessage?.ask === "command_output") {
-								// If the last ask is a command_output, and we
+							if (secondLastMessage?.ask === "command") {
+								// If the last ask is a command, and we
 								// receive an api_req_started, then that means
 								// the command has finished and we don't need
 								// input from the user anymore (in every other
@@ -304,12 +297,18 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 								setEnableButtons(false)
 							}
 							break
+						case "command_output":
+							setTextAreaDisabled(false)
+							setClineAsk("command_output")
+							setEnableButtons(true)
+							setPrimaryButtonText(t("chat:proceedWhileRunning.title"))
+							setSecondaryButtonText(t("chat:killCommand.title"))
+							break
 						case "api_req_finished":
 						case "error":
 						case "text":
 						case "browser_action":
 						case "browser_action_result":
-						case "command_output":
 						case "mcp_server_request_started":
 						case "mcp_server_response":
 						case "completion_result":
@@ -399,7 +398,6 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 						case "tool":
 						case "browser_action_launch":
 						case "command": // User can provide feedback to a tool or command use.
-						case "command_output": // User can send input to command stdin.
 						case "use_mcp_server":
 						case "completion_result": // If this happens then the user has feedback for the completion result.
 						case "resume_task":
@@ -441,6 +439,8 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 		(text?: string, images?: string[]) => {
 			const trimmedInput = text?.trim()
 
+			console.log(`handlePrimaryButtonClick -> clineAsk=${clineAsk}`, text)
+
 			switch (clineAsk) {
 				case "api_req_failed":
 				case "command":
@@ -521,6 +521,7 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 					vscode.postMessage({ type: "terminalOperation", terminalOperation: "abort" })
 					break
 			}
+
 			setTextAreaDisabled(true)
 			setClineAsk(undefined)
 			setEnableButtons(false)
@@ -749,7 +750,8 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 				return alwaysAllowExecute && isAllowedCommand(message)
 			}
 
-			// For read/write operations, check if it's outside workspace and if we have permission for that
+			// For read/write operations, check if it's outside workspace and
+			// if we have permission for that.
 			if (message.ask === "tool") {
 				let tool: any = {}
 
@@ -1114,13 +1116,19 @@ const ChatViewComponent: React.ForwardRefRenderFunction<ChatViewRef, ChatViewPro
 		}
 
 		const autoApprove = async () => {
-			if (isAutoApproved(lastMessage)) {
+			if (lastMessage?.ask && isAutoApproved(lastMessage)) {
 				// Add delay for write operations.
-				if (lastMessage?.ask === "tool" && isWriteToolAction(lastMessage)) {
+				if (lastMessage.ask === "tool" && isWriteToolAction(lastMessage)) {
 					await new Promise((resolve) => setTimeout(resolve, writeDelayMs))
 				}
 
-				handlePrimaryButtonClick()
+				vscode.postMessage({ type: "askResponse", askResponse: "yesButtonClicked" })
+
+				setInputValue("")
+				setSelectedImages([])
+				setTextAreaDisabled(true)
+				setClineAsk(undefined)
+				setEnableButtons(false)
 			}
 		}
 		autoApprove()

webview-ui/src/components/chat/CommandExecution.tsx

@@ -14,7 +14,7 @@ import { cn } from "@src/lib/utils"
 import { Button } from "@src/components/ui"
 
 interface CommandExecutionProps {
-	executionId?: string
+	executionId: string
 	text?: string
 }
 
@@ -36,10 +36,6 @@ export const CommandExecution = ({ executionId, text }: CommandExecutionProps) =
 
 	const onMessage = useCallback(
 		(event: MessageEvent) => {
-			if (!executionId) {
-				return
-			}
-
 			const message: ExtensionMessage = event.data
 
 			if (message.type === "commandExecutionStatus") {