You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@maka.apache.org by as...@apache.org on 2026/08/27 17:00:00 UTC
(maka) branch main updated: test(desktop): story-cover the interrupted tool row after a turn abort (#3951)
This is an automated email from the ASF dual-hosted git repository.
Astro-Han pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/maka.git
The following commit(s) were added to refs/heads/main by this push:
new 4e1678254 test(desktop): story-cover the interrupted tool row after a turn abort (#3951)
4e1678254 is described below
commit 4e16782540eca8af28a9bd5a023ecd9a1b8e7e06
Author: Xiao Liu <li...@qq.com>
AuthorDate: Fri Aug 28 00:59:56 2026 +0800
test(desktop): story-cover the interrupted tool row after a turn abort (#3951)
* test(desktop): story-cover the interrupted tool row after a turn abort
The interrupted tool state — a tool the user stops mid-run — rendered in
no story. Add InterruptedToolAfterTurnAbort beside RunningStatusDuringToolRun:
a Bash tool_call whose abort settles as a cancelled `terminal` result
(exitCode 130, "Command cancelled"), so `toolResultActivityStatus` derives
the row as `interrupted` — there is no interrupted turn status; the state
comes from the settled result, and the aborted turn retains that partial
output. It projects through the real ChatView -> materializeTurns ->
ToolTrow path, rendering inside the production `.maka-turn` frame; the
session is aborted with its timestamp aligned to the transcript.
The result passes the canonical shell-result decoder, and the ledger
carries the fields the AI-SDK backend persists (assistant step + stepId,
origin, modelVisibility, durationMs). running/completed/errored already
have coverage; this adds only the genuinely missing interrupted row.
First surface under #3944.
Refs #3944, #3893
Generated-by: Claude Code
* test(desktop): make the interrupted-tool fixture reachable and faithful
Address review of InterruptedToolAfterTurnAbort. A `tsc` typecheck settles
in under a second, so a multi-second cancel can't happen; and hand-written
command output isn't the bytes the command produces. Use `npm test`, which
runs for minutes (build:test then the runner), cancelled at ~16s while still
building — genuinely `cancelled`/130, inside the 120s foreground default so
it is not `timed_out`/124.
The retained stdout is a verbatim prefix of a real `npm test` run killed
mid-build, so the expanded panel shows the exact bytes — leading/trailing
newlines included, no test-runner interleaving, nothing truncated. It
derives through the real ChatView -> materializeTurns -> ToolTrow path and
renders in the production `.maka-turn` frame; the session is aborted with
its timestamp aligned to the transcript.
Verified: shell-result decoder -> valid; materializeTurns -> tool.status
interrupted, turn.status aborted; storybook smoke green (196 stories).
Refs #3944, #3893
Generated-by: Claude Code
---
apps/desktop/stories/app-shell.stories.tsx | 103 +++++++++++++++++++++++++++++
1 file changed, 103 insertions(+)
diff --git a/apps/desktop/stories/app-shell.stories.tsx b/apps/desktop/stories/app-shell.stories.tsx
index 340014f60..686d364a3 100644
--- a/apps/desktop/stories/app-shell.stories.tsx
+++ b/apps/desktop/stories/app-shell.stories.tsx
@@ -573,6 +573,109 @@ export const RunningStatusDuringToolRun: Story = {
),
};
+// A real prefix of `npm test` stdout, copied verbatim from an actual run killed
+// mid-build (the full suite runs for minutes, so a cancel here is genuinely
+// reachable). Kept short by cutting inside the build phase — no test-runner
+// interleaving, no truncation. Cutting the fixture from a real run is what keeps
+// the expanded panel's bytes honest.
+const NPM_TEST_STDOUT_AT_CANCEL = "\n> maka@0.2.0 test\n> npm run build:test && node scripts/run-workspace-tests-parallel.mjs --concurrency=3\n\n\n> maka@0.2.0 build:test\n> npm run clean && npm --workspace @maka/core run build && npm --workspace @maka/storage run build && npm --workspace @maka/mcp run build && npm --workspace @maka/runtime run build && npm --workspace @maka/runtime-host run build && npm --workspace @maka/computer-use run build && npm --workspace @maka/eval run build && [...]
+
+// Real path: run the full test suite → the user hits stop before it returns.
+// Aborting settles the call as a cancelled `terminal` result (isError), and
+// `toolResultActivityStatus` maps a cancelled terminal to `interrupted`. There is
+// no `interrupted` turn status (only running/completed/aborted/failed) — the
+// tool-level state is derived from the settled result, not asserted. Because the
+// turn kept that partial result, `partialOutputRetained` is true.
+//
+// `npm test` runs for minutes (build:test then the runner), so a cancel at ~16s is
+// still inside a running process — it settles `cancelled`/130, not `timed_out`/124
+// (which needs the 120s foreground default) and not a `completed` run. The retained
+// stdout is a verbatim prefix of a real run (see NPM_TEST_STDOUT_AT_CANCEL), cut in
+// the build phase so there is no runner interleaving and nothing is truncated.
+//
+// This is the interrupted counterpart to RunningStatusDuringToolRun, and the only
+// story that reaches the interrupted tool row. It goes through the real
+// ChatView → materializeTurns → ToolTrow path, so the row renders inside the
+// production `.maka-turn` frame. The session is `aborted` too, so the sidebar row
+// and composer agree with the transcript instead of still reading as active.
+export const InterruptedToolAfterTurnAbort: Story = {
+ render: () => (
+ <ComposedShell
+ session={{ status: 'aborted', lastMessageAt: NOW - 98_000 }}
+ chat={{
+ messages: [
+ user('msg-i-1', 'turn-i', 2, '把整套测试跑一遍，我刚改完 core，想确认没打断别的。'),
+ {
+ type: 'turn_state',
+ id: 'state-i-running',
+ turnId: 'turn-i',
+ ts: NOW - 118_000,
+ status: 'running',
+ partialOutputRetained: false,
+ },
+ {
+ type: 'assistant',
+ id: 'msg-assistant-i',
+ turnId: 'turn-i',
+ ts: NOW - 116_000,
+ text: '跑 npm test —— 先全量重建再跑用例。',
+ modelId: 'claude-sonnet-4-5',
+ },
+ {
+ type: 'tool_call',
+ id: 'tool-i-1',
+ turnId: 'turn-i',
+ ts: NOW - 114_000,
+ toolName: 'Bash',
+ activityKind: 'command',
+ stepId: 'msg-assistant-i',
+ origin: 'provider',
+ modelVisibility: 'visible',
+ args: { command: 'npm test' },
+ },
+ {
+ type: 'tool_result',
+ id: 'tool-i-1-result',
+ turnId: 'turn-i',
+ ts: NOW - 98_000,
+ toolUseId: 'tool-i-1',
+ isError: true,
+ durationMs: 16_000,
+ origin: 'provider',
+ modelVisibility: 'visible',
+ content: {
+ kind: 'terminal',
+ cwd: '/workspace/maka-agent/.worktree/storybook',
+ cmd: 'npm test',
+ status: 'cancelled',
+ exitCode: 130,
+ failureMessage: 'Command cancelled',
+ output: {
+ mode: 'pipes',
+ stdout: NPM_TEST_STDOUT_AT_CANCEL,
+ stderr: '',
+ stdoutTruncated: false,
+ stderrTruncated: false,
+ redacted: false,
+ },
+ },
+ },
+ {
+ type: 'turn_state',
+ id: 'state-i-aborted',
+ turnId: 'turn-i',
+ ts: NOW - 98_000,
+ status: 'aborted',
+ abortedAt: NOW - 98_000,
+ abortSource: 'renderer.stop_button',
+ partialOutputRetained: true,
+ },
+ ],
+ }}
+ />
+ ),
+};
+
// Real path: Desktop Computer Use is exposed through the Runtime Host Client
// Capability bridge. The settled observation establishes the confirmed target;
// the following sequence inherits it while live progress replaces the generic
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@maka.apache.org
For additional commands, e-mail: commits-help@maka.apache.org