Repository navigation
fix(network): report wire requests and retain response evidence - #758
Open
lanyue-llk wants to merge 1 commit into
Open
lanyue-llk wants to merge 1 commit into
lanyue-llk wants to merge 1 commit into
Conversation
lanyue-llk
force-pushed
the
codex/pr640-reopen-layout
branch
from
September 22, 2026 12:44
7bc4a0a to
ef06d30
Compare
lanyue-llk
force-pushed
the
codex/pr640-reopen-network
branch
from
September 22, 2026 12:44
1115a24 to
85f37de
Compare
This was referenced Sep 22, 2026
Sequential Navigation Soak A/B✅ HEAD completed the 200-navigation resilience and memory observation. One browser process, one target, and one CDP session navigate CSDN → SegmentFault → Huaban → example.com repeatedly. This run issued 200 Common ancestor Session resilience
A public-page failure is reported separately from an unrecoverable session. The soak requires all navigation attempts, zero failed recovery, zero lifecycle/network ordering violations, and complete resource evidence. Process-tree memory
HEAD memory by 50-navigation quarter
Raw reports and per-navigation boundary samples: workflow run and full Public-site timing and memory are observational. Use the A/B deltas and trend shape as evidence, not as a deterministic performance threshold. |
Spider Bench A/B✅ All benchmark browser service runs completed; results are informational. Public HEAD: 184 / 240 rows; 38 / 48 sites produced rows; 10 unexpected empty sites. ✅ Deterministic fixture contract is clean. Common ancestor Public 48-site run · informational
Public-site content, timing, and memory are noisy. This single A/B run reports evidence only; site outcome counts explain missing rows without treating them as a deterministic regression. CPU and memory timelinesThese are bounded 40-point views of the same complete process-tree samples used by the HTML report. CPU uses 100% per occupied logical core; memory lines are RSS first and PSS second when complete PSS samples are available. Basexychart-beta
title "Base CPU"
x-axis "Elapsed seconds" [1.037, 2.039, 2.539, 3.541, 4.543, 6.043, 7.045, 8.047, 9.047, 10.549, 11.549, 12.551, 14.054, 15.055, 16.055, 17.558, 18.56, 19.56, 21.063, 22.064, 22.565, 23.065, 24.065, 25.567, 26.568, 27.57, 29.071, 30.072, 31.072, 32.575, 33.576, 34.577, 35.578, 37.081, 38.082, 39.084, 40.586, 41.087]
y-axis "CPU percent" 0 --> 500
line [9.98, 121.7, 127.89, 141.77, 127.68, 210.07, 143.6, 157.55, 150.03, 147.62, 128, 153.6, 129.68, 87.98, 149.94, 73.86, 163.7, 88.03, 143.68, 97.92, 109.97, 109.94, 124.08, 13.96, 147.9, 109.86, 103.92, 71.96, 33.99, 161.73, 93.83, 23.96, 4, 131.59, 207.81, 51.9, 127.86, 217.48]
xychart-beta
title "Base memory: RSS then PSS"
x-axis "Elapsed seconds" [0.036, 1.037, 2.039, 2.539, 3.541, 4.543, 6.043, 7.045, 8.047, 9.047, 10.549, 11.549, 12.551, 14.054, 15.055, 16.055, 17.558, 18.56, 19.56, 21.063, 22.064, 22.565, 23.065, 24.065, 25.567, 26.568, 27.57, 29.071, 30.072, 31.072, 32.575, 33.576, 34.577, 35.578, 37.081, 38.082, 39.084, 40.586, 41.087, 41.57]
y-axis "Memory MiB" 0 --> 500
line [12.25, 68.23, 113.84, 123.02, 129.9, 128.27, 179.15, 207.04, 177.24, 181.07, 180.11, 195.54, 187.76, 169.04, 160.26, 167.34, 154.74, 178.45, 195.05, 179.67, 267.89, 269.27, 198.98, 227.23, 161.92, 182.15, 197.78, 178.71, 173.3, 168.82, 207.35, 191.05, 182.57, 169.95, 197.08, 209.98, 170.4, 185.77, 208.18, 0]
line [9.08, 64.71, 109.78, 118.65, 126.62, 124.75, 175.64, 201.63, 173.75, 177.55, 176.6, 192.09, 184.31, 165.53, 156.75, 163.82, 151.25, 175.49, 189.59, 176.19, 264.39, 265.33, 195.45, 223.76, 158.41, 178.64, 194.25, 175.21, 169.72, 165.31, 204.05, 187.6, 179.06, 166.44, 193.59, 206.22, 166.93, 182.38, 204.75, 0]
HEADxychart-beta
title "HEAD CPU"
x-axis "Elapsed seconds" [1.535, 2.537, 4.04, 5.54, 6.541, 8.043, 9.545, 11.047, 12.049, 13.551, 15.052, 16.053, 17.556, 19.057, 20.058, 21.56, 23.06, 23.562, 24.562, 25.564, 27.066, 28.568, 29.068, 29.568, 31.07, 32.572, 33.573, 35.074, 36.576, 37.577, 39.079, 40.581, 42.082, 43.085, 44.586, 46.088, 47.089, 48.09]
y-axis "CPU percent" 0 --> 500
line [123.92, 127.8, 113.76, 179.94, 151.75, 95.97, 181.93, 77.8, 131.83, 139.92, 133.94, 31.94, 15.96, 81.97, 105.74, 6, 6, 0, 73.95, 69.89, 111.94, 99.9, 110.01, 110.01, 125.92, 125.84, 145.88, 158.07, 83.88, 49.94, 115.89, 91.88, 7.99, 7.98, 121.82, 33.96, 131.88, 221.63]
xychart-beta
title "HEAD memory: RSS then PSS"
x-axis "Elapsed seconds" [0.035, 1.535, 2.537, 4.04, 5.54, 6.541, 8.043, 9.545, 11.047, 12.049, 13.551, 15.052, 16.053, 17.556, 19.057, 20.058, 21.56, 23.06, 23.562, 24.562, 25.564, 27.066, 28.568, 29.068, 29.568, 31.07, 32.572, 33.573, 35.074, 36.576, 37.577, 39.079, 40.581, 42.082, 43.085, 44.586, 46.088, 47.089, 48.09, 48.547]
y-axis "Memory MiB" 0 --> 500
line [12.33, 114.1, 137.55, 140.08, 153.63, 173.51, 165.76, 193.36, 163.85, 189.07, 185.14, 161.28, 144.61, 148.8, 143.65, 158.04, 143.72, 145.88, 145.88, 151.84, 172.79, 171.3, 259.63, 289.59, 199.71, 232.47, 172.57, 176.62, 167.59, 168.56, 160.74, 199.84, 180.29, 166.84, 166.02, 179.93, 156.17, 170.86, 194.4, 0]
line [9.47, 110.85, 134.35, 136.93, 150.43, 170.48, 162.51, 190.03, 160.41, 185.66, 182.8, 157.88, 141.17, 145.37, 140.64, 154.57, 140.27, 142.4, 142.4, 148.38, 169.31, 168.53, 256.26, 286.18, 196.82, 229.06, 168.85, 173.4, 164.16, 165.19, 157.3, 196.52, 177.1, 163.41, 162.58, 176.44, 152.53, 167.37, 190.89, 0]
Deterministic fixture · required contractThe fixture exercises 15 routes: 8 are expected to emit rows and 7 intentionally exercise empty, loading, or timeout behavior.
Full HTML, JSON, CSV, logs, and page snapshots: workflow run and full The public-web A/B is informational. The exact HEAD fixture contract runs as a separate required CI check; timing and public-site content deltas remain non-blocking. |
CI Regression ReportSource CI run · source state at render:
Release regression — ✅ HEAD/base failures 0/0; raw binary +0.031838%Package and image size
Startup latency and PSS
Concurrency matrix
Frontend differential — ✅ 1,020/1,020 cases matched; 0 issues
Agent episodes · Moli vs Chromium — ✅ 8/8 Moli episodes passed; 0 failures
Runtime and CDP session contracts — ✅ 26 contract cases; 0 failures
CDP smoke — ✅ 52/52 groups passed; 616 scenariosWorkers: 4 · cumulative group time: 136.01 s · failed groups: 0
WebMainBench · 545 pages — ✅ 545/545 completed; 0 unexpected failures
Frozen HTML · Linux · Diagnostics: All artifact fields are parsed by the trusted default-branch renderer; missing or invalid inputs remain visible as unavailable. |
lanyue-llk
force-pushed
the
codex/pr640-reopen-layout
branch
from
September 24, 2026 04:22
ef06d30 to
d979e94
Compare
lanyue-llk
force-pushed
the
codex/pr640-reopen-network
branch
from
September 24, 2026 04:22
85f37de to
d7c049b
Compare
lanyue-llk
force-pushed
the
codex/pr640-reopen-layout
branch
from
September 24, 2026 05:12
d979e94 to
f4461d1
Compare
lanyue-llk
force-pushed
the
codex/pr640-reopen-network
branch
from
September 24, 2026 05:12
d7c049b to
568b7b1
Compare
Contributor
Author
lanyue-llk
force-pushed
the
codex/pr640-reopen-network
branch
from
September 29, 2026 12:35
568b7b1 to
a32dbf2
Compare
Contributor
Author
lanyue-llk
force-pushed
the
codex/pr640-reopen-network
branch
from
October 8, 2026 03:56
a32dbf2 to
6d7c57c
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
请求实际成功时，CDP 仍可能报告错误的逐跳方法、缺失的实际请求头，或在导航后丢失响应体。发送前的请求描述不等于传输最终发送的数据。本 PR 将网络事件与响应证据归属到实际交换，并为跨导航响应体提供显式、有界、按会话隔离的保存机制。
一图看懂
图示说明问题、修复机制与行为边界；点击可查看原图。验收归属不等于当前提交通过结果，版本和测试证据见下文。
场景、根因与预期
机制与代码阅读路线
moli-fetch/src/network_fetch_result.rs和request.rs：wire 请求行及最终头的观察，共享重定向方法与 body-header 规则。先核对 Fetch 的逐跳语义，再看协议投影。moli-renderer-v8/src/network/navigation/loader.rs与 child documents loads：主/子文档都携带实际网络元数据。moli-protocol/src/domains/network/main_document_progress/、redirect_request.rs、events.rs、output_queue.rs：按交换归属日志；保留普通事件与 ExtraInfo 的不同含义，不能用 staged cookie query 覆盖已经观察到的 wire 头。network/agent.rs、settings.rs、connection 的 network owner/runtime slot：有界 body owner 管保存，session claim 管可见性。导航、关闭、撤销、缩小预算必须同步 owner 和索引。moli-bounded-buffer/src/buffer.rs：复用 FIFO、retain、限额收缩等通用原语。runtime/protocol_output/transport_memory.rs计入方法、头、journal 与相关消息的内存；零字节 body 也不能绕过预算。这 28 个文件跨越传输、renderer、协议和公共缓冲区，是同一证据链的不同边界。较大 agent 改动应围绕所有权、会话权限和预算不变量审查；不增加网站特例，也不由 CDP 猜测网络实际上发送了什么。
回归用例与反例
network/tests/redirect_wire.rs::form_post_redirect_chain_cdp_metadata_matches_wire_requests：真实服务端收到的请求与 CDP 方法/头逐跳对照redirect_request.rs：redirect_status_method_matrix_matches_fetch_semantics、redirected_get_does_not_recover_original_post_on_later_307main_document_progress/tests.rs：不能补回删除的头，不能猜截断日志的跳数，升级请求的 wire 头归到正确交换page/producer_tests.rs::child_document_network_reports_actual_redirect_methods_and_request_headersnetwork/tests/response_body.rs：显式启用且正预算才保存；disable/detach/close/其他 target 不泄漏Case 对应与验收版本
本 PR 负责
339,595,676。原第 10 类缺失 Referer 对应339,676；595 原属第 2 类链接几何，最终需要网络证据与 #756 几何组合验收，不能把其成功全部归因于本 PR。6d7c57c395505c84eebedfb51d2ab74d821e8ae4已在 mainb21b4f05615c1b9aa71ea2074084646895d82be7上真正 rebase，保留单提交。CI 37725108193 19/19 成功。a32dbf25671a4ac0cf2a0bff8139a40d13f02912的官方任务339,676评分 1.0；595 的历史组合验收为 16/16 动作、默认 30 秒、评分 1.0，结束恢复订阅状态。这些不是当前 HEAD 的官方任务复跑。15faebd1683b94fb4aa646f3ffe695a2bb45a44b在原冻结合同下 2/3，676 失败；显式就绪 v2 合同下 3/3、零重试。报告保留就绪合同和中间失败，不能把不同合同合并成原合同全部通过。网络历史判定。当前 base 是 main；595 的运行场景仍依赖 #756 的布局行为。来源：原网络 PR #737。
Reviewer 复现：真实请求与 3 个 badcase
先在本 PR 的 Moli checkout 跑真实 HTTP 重定向与响应体测试，不需要 WebArena：
cargo nextest run --no-fail-fast -p moli-protocol \ -E 'test(form_post_redirect_chain_cdp_metadata_matches_wire_requests) | test(durable_response_body_) | test(observed_request_headers_do_not_restore_a_referrer_removed_by_transport_policy)'预期：真实服务端与 CDP 都是 POST → GET → GET；不会恢复已删除的 Referer。跨导航 body 保存需显式开启和正预算，disable/detach/其他 target 不泄漏。
完整复现指南与环境准备 · 下载复现包。包内有按 case 选题的脚本、固定动作和 DOM 采集合同，不含历史答案。代码复用 browser-eval 的执行、HAR 采集与评分器，不是重新实现评分。
首次使用：解压到 Moli 仓库外，将
REPRO_DIR设为解压目录的绝对路径。独立克隆 browser-eval 后，固定版本并安装锁定依赖：将
site-map.json的主机和端口换成自己使用的 WebArena 测试站点。需要 Reddit、GitLab、Shopping、Shopping Admin；不用 Map/Wikipedia。镜像、初始化、测试账号和地址要求见指南。下文命令都在这个 browser-eval 目录执行。脚本自行启动待审二进制，使用新 profile 和可用端口，结束后只停止它启动的进程。先跑 339、676：
595 需同时具备 #756 的布局和 #758 的网络修改，单个 PR 的二进制不能代表组合结果。用包含两项修改的二进制运行：
595 会修改订阅，运行前确认未订阅，结束后取消订阅或重置自己的环境。不能在其他人使用的站点跑写入或重置。676 若网格尚未加载，应先看 trace 判断就绪问题；本包不偷偷增加 readiness v2 或延长超时，也不把历史 v2 的 3/3 当成原合同结果。
保存和判断结果
每轮使用新结果目录。
inputs.json记录二进制、输入哈希及回放/评分器版本；每题保存result.json、trace.json、network.har、实际响应与评分。动作失败时评分文件可能尚未生成，先看failure_stage和 trace；网络问题对照 HAR 的方法、URL、Referer、重定向与状态码。默认每动作 30 秒、每题 300 秒，每题只执行一次。比较历史失败时用 1.1.9 官方二进制和待审 PR，并保持相同站点初始数据。当前 main 已包含其他修复，不应预设所有历史题仍会在 main 上失败。
Reviewer 重点
逐跳方法是否同时对照服务端真实请求？观测不完整时是否明确回退？会话撤销、导航及预算收缩后，可见性、内存计费和索引是否保持一致？main 的 no-cors 非 follow 重定向拒绝语义不能被观测路径改写。
整组验收与证据入口
这组修复来自 #640，对应 1.1.9 的 34 个失败 case，加上 1.1.8 已失败的 386，共 35 个唯一 case。按最终职责分配为：#756 布局 27、#757 输入 4、#758 网络 3、#761 文本 1；#755、#759、#760 以各自独立功能契约验收，不重复认领这 35 个 case。