fix(e2e): integrate Studio testing and owner approval workflows

Cargo.toml

@@ -387,10 +387,7 @@ used_underscore_binding = "allow"
 ## sandbox dirs, mobile-layout detection, ArkUI native inputs), with the legacy
 ## `old/` crate copies removed so the repo resolves as a cargo git dependency.
 [patch."https://github.com/makepad/makepad"]
-## Pinned to ZhangHanDong/makepad dev at the merge of PR #6 (Splash-isolate
-## shader identity, headless renderer correctness, ttf-parser gvar-alloc for
-## Apple variable fonts, the CoreText outline fallback for hvgl-only system
-## fonts like PingFang, and the closed-Modal overdraw fix), plus the
-## selectable-Html fix from PR #5.
-makepad-widgets = { git = "https://github.com/ZhangHanDong/makepad", rev = "1dd8d372cfb65b20160a3782586ed4d593cf8193", features = ["serde"] }
-makepad-code-editor = { git = "https://github.com/ZhangHanDong/makepad", rev = "1dd8d372cfb65b20160a3782586ed4d593cf8193" }
+## Isolated E2E integration of reviewed Dock visibility PR 8.
+## Build Makepad Studio from this same commit for GUI E2E tests.
+makepad-widgets = { git = "https://github.com/ZhangHanDong/makepad", rev = "45e826c2814fccd2c69d558b27775858444c907d" }
+makepad-code-editor = { git = "https://github.com/ZhangHanDong/makepad", rev = "45e826c2814fccd2c69d558b27775858444c907d" }

proptest-regressions/event_preview.txt

@@ -0,0 +1,7 @@
+# Seeds for failure cases proptest has generated in the past. It is
+# automatically read and these particular cases re-run before any
+# novel cases are generated.
+#
+# It is recommended to check this file in to source control so that
+# everyone who runs the test benefits from these saved cases.
+cc 0c081b28fcee5c8c4b5439561a52585269c1cfb4d75510d9001f584fb8ac3aea # shrinks to body = "", detail = ""

resources/i18n/en.json

@@ -686,6 +686,7 @@
   "forward_modal.button.forward": "Forward",
   "forward_modal.popup.submitting": "Forwarding message...",
   "room_screen.popup.action_response.failed": "Failed to send action response.\n\nError: {error}",
+  "room_screen.popup.action_response.outcome_unknown": "The send outcome is unknown. Check the canonical approval status before trying again.\n\nError: {error}",
   "room_screen.send_state.resend_modal.title": "Message not sent",
   "room_screen.send_state.resend_modal.body": "This message could not be sent. Resend it?",
   "room_screen.send_state.resend_modal.body_reason": "This message could not be sent: {reason}. Resend it?",
@@ -694,6 +695,17 @@
   "room_screen.popup.approval_expired": "This approval request has expired. Ask the agent to create a new request.",
   "room_screen.approval.pending": "Pending",
   "room_screen.approval.expired": "Expired",
+  "room_screen.approval.approved": "Approved",
+  "room_screen.approval.denied": "Denied",
+  "room_screen.approval.consumed": "Consumed",
+  "room_screen.approval.unavailable": "Unavailable",
+  "room_screen.approval.submitting": "Submitting…",
+  "room_screen.approval.sent_awaiting_canonical": "Sent — awaiting status",
+  "room_screen.approval.outcome_unknown": "Send outcome unknown",
+  "room_screen.approval.task_result_unknown": "Approval consumption does not confirm that the task completed.",
+  "room_screen.approval.view_pending": "View pending approvals",
+  "room_screen.approval.navigation_unavailable": "The approval room is not available yet. Wait for room sync, then try again.",
+  "room_screen.approval.navigation_ambiguous": "More than one approval room matches this agent. Ask its owner to update the room binding.",
   "room_screen.modal.delete_message.title": "Delete Message",
   "room_screen.modal.delete_message.body": "Are you sure you want to delete this message? This cannot be undone.",
   "room_screen.modal.delete_message.accept": "Delete",

resources/i18n/zh-CN.json

@@ -686,6 +686,7 @@
   "forward_modal.button.forward": "转发",
   "forward_modal.popup.submitting": "正在转发消息...",
   "room_screen.popup.action_response.failed": "发送动作响应失败。\n\n错误：{error}",
+  "room_screen.popup.action_response.outcome_unknown": "发送结果未知。再次尝试前，请先检查 canonical 审批状态。\n\n错误：{error}",
   "room_screen.send_state.resend_modal.title": "消息未发送",
   "room_screen.send_state.resend_modal.body": "这条消息未能发出，是否重新发送？",
   "room_screen.send_state.resend_modal.body_reason": "这条消息未能发出：{reason}。是否重新发送？",
@@ -694,6 +695,17 @@
   "room_screen.popup.approval_expired": "该审批请求已过期，请让 agent 重新发起请求。",
   "room_screen.approval.pending": "等待审批",
   "room_screen.approval.expired": "已过期",
+  "room_screen.approval.approved": "已批准",
+  "room_screen.approval.denied": "已拒绝",
+  "room_screen.approval.consumed": "已消费",
+  "room_screen.approval.unavailable": "不可用",
+  "room_screen.approval.submitting": "正在提交…",
+  "room_screen.approval.sent_awaiting_canonical": "已发送，等待状态更新",
+  "room_screen.approval.outcome_unknown": "发送结果未知",
+  "room_screen.approval.task_result_unknown": "审批已消费不代表任务已经完成。",
+  "room_screen.approval.view_pending": "查看待审批",
+  "room_screen.approval.navigation_unavailable": "审批房间暂不可用。请等待房间同步后重试。",
+  "room_screen.approval.navigation_ambiguous": "此 Agent 对应多个审批房间，请联系所有者更新房间绑定。",
   "room_screen.modal.delete_message.title": "删除消息",
   "room_screen.modal.delete_message.body": "确认要删除这条消息吗？此操作无法撤销。",
   "room_screen.modal.delete_message.accept": "删除",

scripts/spec-guard.sh

@@ -12,10 +12,13 @@
 #      `manual_test_*` selectors, which agent-spec reports as `skip`, and the
 #      boundary layer applies the whole change set to every spec, so a naive
 #      `agent-spec guard` fails on every PR. Instead:
-#        - specs changed in this change set are the active contracts: run
-#          `lifecycle` WITH the change set (boundaries + tests), fail on failed>0;
+#        - a single changed contract receives the full change set;
+#        - multiple active contracts collectively cover every changed path,
+#          then each runs `lifecycle` with its owned paths (boundaries + tests).
+#          Shared files go to every declaring task; explicit denies still fail;
 #        - all other specs are regression checks: run `verify` WITHOUT a change
-#          set (tests only), fail on failed>0. Skips are tolerated.
+#          set (tests only), fail on failed>0. Skips and legacy regression
+#          uncertainty remain nonblocking, explicitly reported as UNVERIFIED.
 #
 # Usage:
 #   scripts/spec-guard.sh [--change-scope staged|worktree] [--base <git-ref>] [--fast]
@@ -101,39 +104,17 @@ for cap in specs/capabilities/*.spec.md; do
 done
 
 # ---- 4. task specs ---------------------------------------------------------------
-# 4a. Changed specs are the active contracts of this change: verify them WITH the
-#     change set (boundaries + bound tests). Manual `skip`s are tolerated; any
-#     `failed > 0` (test or boundary violation) fails the gate.
+# 4a. Changed specs own the change collectively, not by intersecting unrelated
+#     allow-lists. The helper checks total path coverage and explicit denies,
+#     then runs the native lifecycle boundary verifier and tests for each scope.
+#     Manual skips stay visible; failed/uncertain active contracts fail the gate.
 # 4b. Every other spec is a regression check: verify WITHOUT a change set (no
 #     boundary layer, since foreign files would trivially violate them) and fail
 #     only on `failed > 0`.
 say "4/4 task specs: changed = contract check (with change set), others = regression (tests only)"
 CHANGE_ARGS=()
 for f in "${CHANGED[@]:-}"; do [ -n "$f" ] && [ -e "$f" ] && CHANGE_ARGS+=(--change "$f"); done
-summary_failed() { grep -Eo '"failed": [0-9]+' | head -1 | grep -Eo '[0-9]+'; }
-is_changed_spec() { local x; for x in "${CHANGED_SPECS[@]:-}"; do [ "$x" = "$1" ] && return 0; done; return 1; }
-for spec in specs/*.spec.md; do
-  case "$spec" in specs/project.spec.md) continue;; esac
-  if is_changed_spec "$spec"; then
-    out="$(agent-spec lifecycle "$spec" --code . --run-log-dir .agent-spec/runs "${CHANGE_ARGS[@]}" --format json 2>&1 || true)"
-    n="$(echo "$out" | summary_failed || echo "?")"
-    if [ "$n" != "0" ]; then
-      echo "FAIL (contract): $spec — failed=$n"
-      echo "$out" | grep -E '"reason": "not covered|"verdict": "fail"' -B3 | grep -E 'reason|step_text|scenario_name' | head -12
-      fail=1
-    else
-      echo "ok (contract): $spec"
-    fi
-  else
-    out="$(agent-spec verify "$spec" --code . --format json 2>&1 || true)"
-    n="$(echo "$out" | summary_failed || echo "?")"
-    if [ "$n" != "0" ]; then
-      echo "FAIL (regression): $spec — failed=$n"
-      echo "$out" | grep -E '"verdict": "fail"' -B8 | grep -E 'scenario_name|reason' | head -8
-      fail=1
-    fi
-  fi
-done
+if ! python3 scripts/spec_guard.py "${CHANGE_ARGS[@]}"; then fail=1; fi
 
 if [ "$fail" = "0" ]; then echo; echo "spec-guard: OK"; else echo; echo "spec-guard: FAILED"; fi
 exit "$fail"

scripts/spec_guard.py

@@ -0,0 +1,149 @@
+"""Task composition for spec-guard.sh. Native agent-spec still verifies each scope."""
+import argparse
+from functools import lru_cache
+import json
+from pathlib import Path
+import re
+import subprocess
+import sys
+
+
+def path_pattern(text):
+    # Match the path forms recognized by agent-spec; prose is not an allow-list.
+    text = text.strip()
+    if not (any(c in text for c in '/\\*') or text.endswith(('.rs', '.ts', '.js', '.py', '.md', '.spec'))):
+        return None
+    text = text.strip('`').replace('\\', '/')
+    while text.startswith('./'):
+        text = text[2:]
+    return text.strip('/')
+
+
+def matches(pattern, path):
+    pattern = tuple(p for p in pattern.split('/') if p)
+    path = tuple(p for p in path.split('/') if p)
+
+    @lru_cache(None)
+    def visit(i, j):
+        if i == len(pattern):
+            return j == len(path)
+        if pattern[i] == '**':
+            return visit(i + 1, j) or (j < len(path) and visit(i, j + 1))
+        return (j < len(path)
+            and re.fullmatch(re.escape(pattern[i]).replace(r'\*', '.*'), path[j]) is not None
+            and visit(i + 1, j + 1))
+
+    return visit(0, 0)
+
+
+def partition_changes(changes, documents):
+    """All owners receive shared files; unowned paths and explicit denies fail."""
+    scopes = {spec: [] for spec in documents}
+    rules = {}
+    for spec, document in documents.items():
+        allow, deny = [], []
+        for section in document['sections']:
+            if section['kind'] != 'boundaries':
+                continue
+            for item in section['items']:
+                pattern = path_pattern(item['text'])
+                if pattern and item['category'] in ('allow', 'deny', 'general'):
+                    (allow if item['category'] == 'allow' else deny).append(pattern)
+        rules[spec] = (allow, deny)
+    errors = []
+    for change in changes:
+        owned = False
+        for spec, (allow, deny) in rules.items():
+            if any(matches(pattern, change) for pattern in deny):
+                errors.append(f'{change}: forbidden by {spec}')
+            if any(matches(pattern, change) for pattern in allow):
+                scopes[spec].append(change)
+                owned = True
+        if not owned:
+            errors.append(f'{change}: not covered by any active task boundary')
+    if errors:
+        raise ValueError('\n'.join(errors))
+    return scopes
+
+
+def cli_json(arguments):
+    result = subprocess.run(['agent-spec', *arguments], text=True, capture_output=True)
+    try:
+        document = json.loads(result.stdout)
+    except (ValueError, TypeError) as error:
+        raise ValueError(f'invalid agent-spec JSON (exit {result.returncode}): '
+            f'{(result.stderr or result.stdout)[:1200]}') from error
+    if not isinstance(document, dict):
+        raise ValueError('agent-spec report must be an object')
+    return result.returncode, document
+
+
+def check_report(code, document, regression=False):
+    report = document.get('verification', document)
+    summary = report['summary']
+    counts = {key: summary[key] for key in ('total', 'passed', 'failed', 'skipped', 'uncertain')}
+    if any(type(value) is not int or value < 0 for value in counts.values()):
+        raise ValueError('invalid verification counts')
+    # Preserve the existing nonblocking uncertainty policy for unchanged tasks.
+    # Those results remain UNVERIFIED; active contracts cannot use this policy.
+    permitted_uncertain = regression and counts['uncertain'] > 0
+    unverified = counts['skipped'] > 0 or counts['uncertain'] > 0
+    ok = (code in (0, 1) and counts['total'] > 0 and counts['failed'] == 0
+        and (counts['uncertain'] == 0 or permitted_uncertain)
+        and (code == 0 or counts['skipped'] > 0 or permitted_uncertain))
+    detail = ', '.join(f'{key}={counts[key]}' for key in ('passed', 'failed', 'skipped', 'uncertain'))
+    if not ok or counts['uncertain'] > 0:
+        # Slice parsed diagnostics, never a live stdout pipe. Later tasks still run.
+        failures = [r for r in report.get('results', []) if r.get('verdict') in ('fail', 'uncertain')]
+        for result in failures[:12]:
+            print('  ' + result.get('scenario_name', '<unnamed>'))
+            for step in result.get('step_results', [])[:3]:
+                if step.get('verdict') in ('fail', 'uncertain'):
+                    print(f"    {step.get('step_text', '')}: {step.get('reason', '')}"[:1200])
+    return ok, detail, unverified
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--change', action='append', default=[])
+    args = parser.parse_args()
+    changes = list(dict.fromkeys(args.change))
+    specs = sorted(str(p) for p in Path('specs').glob('*.spec.md') if p.name != 'project.spec.md')
+    active = [spec for spec in specs if spec in changes]
+    scopes = {spec: changes for spec in active}
+    if len(active) > 1:
+        try:
+            documents = {}
+            for spec in active:
+                code, document = cli_json(['parse', spec, '--format', 'json'])
+                if code != 0:
+                    raise ValueError(f'could not parse {spec}: exit {code}')
+                documents[spec] = document
+            scopes = partition_changes(changes, documents)
+        except (ValueError, KeyError, TypeError) as error:
+            print(f'FAIL (composition): {error}', file=sys.stderr)
+            return 1
+    failed = False
+    for spec in specs:
+        changed = spec in scopes
+        mode = 'contract' if changed else 'regression'
+        arguments = ['lifecycle' if changed else 'verify', spec, '--code', '.', '--format', 'json']
+        if changed:
+            arguments += ['--run-log-dir', '.agent-spec/runs']
+            for change in scopes[spec]:
+                arguments += ['--change', change]
+        try:
+            code, document = cli_json(arguments)
+            ok, detail, unverified = check_report(code, document, regression=not changed)
+        except (ValueError, KeyError, TypeError) as error:
+            ok, detail, unverified = False, str(error), False
+        status = ('UNVERIFIED' if unverified else 'ok') if ok else 'FAIL'
+        print(f'{status} ({mode}): {spec} — {detail}', flush=True)
+        if ok and unverified:
+            print('  Nonblocking under existing CI policy; skipped/uncertain scenarios are not passes.', flush=True)
+        failed |= not ok
+    return int(failed)
+
+
+if __name__ == '__main__':
+    sys.exit(main())