Repository navigation
Conversation
…moke The replan-cadence browser smoke still required a fresh machine configuration to render stride strategy `off`. The settled capability default is `evidence_conditioned_v1`: that is what `default_replan_cadence_configuration()` returns, what `tests/capabilities/test_machine_configuration_goal_defaults.py` pins, and what the capability catalog describes as enabled by default. Because the expectation kept the earlier opt-in default, the `dashboard-browser`, `chat-bundle-browser` and `build` lanes failed on `Stride must default off` while the product behaved as intended. Require the settled strategy instead. The remaining `off` assertions in the scenario are explicit opt-outs and the legacy-migration meaning, so they stay unchanged. Validation: node --check examples/personal-workspace-browser/ replan-cadence.mjs. The browser lanes need Playwright and chromium, so the end-to-end run stays with the hosted dashboard lanes. Scope: one assertion, no product default, rendering or migration change. Signed-off-by: ChengLong GU <88359731+gcl-coder@users.noreply.github.com>
|
This pull request has merge conflicts with Choose the remote for the base repository, not an out-of-date fork. git fetch upstream
git rebase upstream/main
# Resolve each conflict, git add the resolved files, then git rebase --continue.
git push --force-with-lease origin HEADFor a same-repository clone whose Keep the DCO |
loopx-agent
left a comment
There was a problem hiding this comment.
Reviewer: model_agent · gpt-6.1-sol · OpenAI · runtime_reported · reasoning_effort=max
动机 / Motivation
维护 dashboard-browser、chat-bundle-browser 和 build 验收场景的开发者。
产品继承默认已经启用 evidence-conditioned Stride,原 smoke 却在新机器设置的首个检查要求 off,导致场景不能继续。
当前修改对齐了首处默认值;同一场景清除 Goal 与机器覆盖后仍要求 off,因此尚未完成恢复这条 smoke 的目标。
不修改产品策略默认值、既有显式 off、旧配置含义或执行权限;不声称已经跑通完整浏览器 lane。
改动思路 / Approach
同一场景的两处产品默认值断言应服从已有配置 owner;保留显式 opt-out 的独立语义,无需改产品或另建测试框架。
当前 PR 的边界是既有 replan-cadence 场景的默认值回归修正,必须覆盖同一场景恢复产品默认的后半段。
最小修复应留在现有场景,不需要改生产默认或增加另一份 smoke。已沿整个场景检查 fresh machine、legacy、显式 Goal off、清除 Goal 覆盖和移除机器配置;“恢复默认”与显式关闭是两种不同输入,不能全局替换所有 off。
具体改动 / Changes
精确 head 72be547b2f76a9c949cf340ddb426c3269432c77,不可变基线 feea5a6e56066eef87889d470d7991922f9f00ec,一文件 +4/-1。独立规范 docs/reference/protocols/goal-vision-replan-contract-v0.md,版本 feea5a6e56066eef87889d470d7991922f9f00ec,Default review cadence 明确:没有 Goal 或已保存机器覆盖时默认 evidence-conditioned;显式 off 和旧 cadence-only 配置保留原义。该准则当前 not_met,首处已改,后半段恢复默认断言未改。
关键内容讲解
- 首个机器默认检查 现在要求 evidence_conditioned_v1,与真实新机器配置和既有默认测试一致;这部分改动合理。
- 同场景移除机器配置 在 Goal 已清除覆盖后 preview/apply remove,再确认 source=capability_default、six effective Turns,属于恢复产品默认,不是 legacy 或显式 opt-out。
- 剩余策略断言 仍要求 englishStrategy 为 off。生产 goalReplanCadenceEditorValue 在 current=null 时直接保留真实 effective 值 evidence_conditioned_v1;该值由同一设置组件作为表单 draft 使用。
对主干的风险 / Risks
[P2] 同步清除覆盖后的产品默认断言。 当前只修首处默认检查,后半段第308行仍以 off 拒绝同一个产品默认。我启动了真实 ChatHTTPServer 的配置 handlers,用隔离存储执行机器 preview/apply、显式 Goal off、清除 Goal 覆盖、移除机器策略,再读取真实 Goal catalog 并调用生产前端转换函数。最终 current=null、source=capability_default、count=6,backend 与前端值均为 evidence_conditioned_v1;剩余 guard 的 off 条件为 false。base/head 相同的产品路径也给出该值,所以这是本次要修的同一默认语义遗漏,不是要求修另一个无关历史失败。
最小修复:同步第308行的产品默认断言,保留显式 Goal off 与 legacy 的负例,并执行完整 LOOPX_PERSONAL_WORKSPACE_SCENARIO=replan-cadence node examples/personal-workspace-browser-smoke.mjs。返回 note 中的 default-off 描述也应准确。这里没有声称实际 DOM 或整条浏览器流程已运行;证据是实际 handler、完整 clear/remove 状态和真实前端转换消费者,非伪造配置返回值。
独立28项机器默认测试、Node syntax 均 PASS。native premerge4direct/6selected 的整体 gate PASS,但保留一项 raw ratchet advisory FAIL;该选择集合没有执行完整浏览器场景,不能覆盖上述遗漏。CQR 应保留这个必要不变量失败,不以语法或 premerge 绿色认证 smoke 已恢复。不读取、等待或轮询远端 CI。
我的整体评价 / Verdict
REQUEST_CHANGES。 当前为 fragmented:同一个用户任务、同一场景的恢复默认路径尚未完整修好。生产行为没有被改,long_horizon=preserved;维护者能否靠这次修正完成整条场景仍为 user_experience=not_yet_proven。现有 smoke 有长期边界保护价值,没有新增重复脚手架或同作者批量 farming 证据;应补齐现有断言,不能删除有效 opt-out/legacy/recovery 覆盖。
已检查变更及其 CLI/配置消费者;没有修改渲染组件、布局或产品控件,视觉评审不适用。
完整浏览器场景与实际 DOM 未运行;真实 HTTP 配置及生产前端转换函数已运行。
剩余第308行仍要求 englishStrategy 为 off;最小修复是对齐该产品默认断言并执行完整 replan-cadence 场景,同时保留 legacy 和显式关闭负例。
English verdict: REQUEST_CHANGES at 72be547b2f76a9c949cf340ddb426c3269432c77. The first default check is corrected, but the same scenario still expects off after removing both overrides. Real HTTP configuration and the production frontend projection return evidence_conditioned_v1. Align the remaining assertion and run the complete scenario; syntax and premerge do not establish that pass.
| if (await strategy.inputValue() !== "evidence_conditioned_v1") | ||
| throw new Error("Machine stride default is not evidence-conditioned"); |
There was a problem hiding this comment.
[P2] 同步清除覆盖后的产品默认断言。 当前只修首处默认检查,后半段第308行仍以 off 拒绝同一个产品默认。我启动了真实 ChatHTTPServer 的配置 handlers,用隔离存储执行机器 preview/apply、显式 Goal off、清除 Goal 覆盖、移除机器策略,再读取真实 Goal catalog 并调用生产前端转换函数。最终 current=null、source=capability_default、count=6,backend 与前端值均为 evidence_conditioned_v1;剩余 guard 的 off 条件为 false。base/head 相同的产品路径也给出该值,所以这是本次要修的同一默认语义遗漏,不是要求修另一个无关历史失败。
|
上游已经自行修复了这一条期望值,本 PR 变成重复提案,因此关闭。 依据:本 PR 只改了 上游提交 两者语义一致,只是错误文案不同。本 PR 的提交 本 PR 没有引入独有覆盖,保留它只会在同一处留下第二份改动和多余说明。关闭后由上游 #7948 的实现统一承担该断言。 旁证: |
|
上游已经自行修复了这一条期望值,本 PR 变成重复提案,因此关闭。 依据:本 PR 只改了 上游提交 两者语义一致,只是错误文案不同。本 PR 的提交 本 PR 没有引入独有覆盖,保留它只会在同一处留下第二份改动和多余说明。关闭后由上游 #7948 的实现统一承担该断言。 旁证: |
1 similar comment
|
上游已经自行修复了这一条期望值,本 PR 变成重复提案,因此关闭。 依据:本 PR 只改了 上游提交 两者语义一致,只是错误文案不同。本 PR 的提交 本 PR 没有引入独有覆盖,保留它只会在同一处留下第二份改动和多余说明。关闭后由上游 #7948 的实现统一承担该断言。 旁证: |
问题
dashboard-browser、chat-bundle-browser和build三条 lane 现在停在同一个断言上:来源是
examples/personal-workspace-browser/replan-cadence.mjs:86:它断言「本机设置 → 能力中心 → Goal 复盘与步幅 → 步幅策略」在一份全新的机器配置上必须显示
off。但产品现在的默认值是
evidence_conditioned_v1:同一份默认值也在
tests/capabilities/test_machine_configuration_goal_defaults.py里被钉住(stride_strategy == "evidence_conditioned_v1"),能力目录的描述也写着 "Evidence-conditioned stride is enabled by default"。也就是说:默认值改成「开启」时,这条 smoke 的期望值没有跟着改,三行 lane 因此长期变红。
改动
只改这一条期望值,并给出与新语义一致的失败文案:
场景里其余的
"off"断言都是显式 opt-out(Goal 级关闭、legacy 配置的既有含义),语义没变,一律不动:goalCadence.current只存{stride_strategy: "off"}—— 显式 Goal 关闭;replanCadenceEditorValue对 legacy 缺省字段仍映射为off—— 迁移前的既有含义;evidence_conditioned_v1—— 与新默认一致。验证
node --check examples/personal-workspace-browser/replan-cadence.mjs→ 通过;default_replan_cadence_configuration()输出见上;同一断言由tests/capabilities/test_machine_configuration_goal_defaults.py覆盖;dashboard-browser/chat-bundle-browser/build三条 lane 验证;本次只把期望值对齐到已生效的产品默认值。范围
一行断言 + 一行文案。不改产品默认值、不改前端渲染、不改 legacy 迁移语义,不改 stride 的执行权限。