Repository navigation
修复 1.6.0 代码审查发现的 25 项问题（下载台账并发 / 频道开关 / 目录误报 / 工程） - #19
Merged
Merged
Conversation
Ledger and downloads (MUC-01, 05, 10, 18, 21, 29): the download ledger is written under a process-wide lock that re-reads and merges, so concurrent download threads, the install command and the check thread no longer overwrite each other's records (9 records reproduced losing to 0); an older download is removed only after the new file is in place; a failed install no longer leaves a truncated jar in mods/ and names the leftover when it cannot clean up; ledger file names are validated before use; the failure message reports the attempts that were actually made; downloads count against the rate limiter. Upstream and checking (MUC-02, 04, 11, 12, 22, 27, 28, 30): release channels are filtered for real (the switch previously had no caller, so beta builds were offered as updates); a negative identification is only cached when Modrinth was actually asked; advisories move inside the try so no request outlives the client; a dead upstream short-circuits instead of searching once per mod; each thread gets its own requests Session; timeouts become (connect, read) with a total deadline for downloads; JSON responses gain a size cap; the wasted per-mod version lookups and the dead assignment are gone. Robustness (MUC-03, 06, 07, 08, 09, 17, 19, 20): a project map saved as GBK/ANSI is reported and ignored instead of taking the whole check down; the MC version vote only counts lower bounds and breaks ties with the real version comparison; an unlistable mods/ returns an explained result rather than an exception; hash reads go through one guarded helper; a malformed cache record is dropped instead of failing every check; backup names are matched with fullmatch; encrypted or corrupt jars no longer break scan_jar's "never raises" promise; metadata extraction has a 4 MiB ceiling. Entry layer (MUC-13, 14, 15, 16, 23, 24, 25, 26, 31): the three silent excepts log; a failed is_server_running probe keeps the last known state instead of muting every notification; taking a staged plan is atomic and single downloads are de-duplicated by file name; a check that is still running when the server stops drops its result; !muc status reads and renders off the command thread; the alias reaches every line that names a command; the cell-width cache is dropped on reload; an over-wide caption is cut to the title bar; numeric config values are clamped to documented ranges with a warning; the module documents its own section order and split plan. Repository and tooling: tools/mutation_check.py replaces the hand-run injection checks (12 mutations, temporary copy, stale anchors reported separately), tools/check_python_floor.py plus a CI job verify the declared 3.8 floor, the bench scripts README-dev cites are in the repository with a path guard test, the packaged changelog is trimmed to one section, and the workflow gains permissions, timeouts, concurrency and a lint job. Verified: 827 tests pass; 12/12 mutations caught; the real MCDR matrix agrees on 2.13.0 / 2.14.1 / 2.15.0 / 2.15.7 / 2.16.0; the artifact is reproducible.
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.
改了什么
修复 v1.6.0 全代码审查（报告：工作区
ServerLogFilter-v1.4.1+ModUpdateChecker-v1.6.0-优化建议.html，原始报告review-batch2/reports/A2、A3、A4）里 ModUpdateChecker 的 25 条发现（4×P1 / 12×P2 / 9×P3 + 优化项），并把仓库级工程项一并补齐。目标版本 v1.6.1。台账与下载
!!muc download的两个后台线程、命令线程上的install与检查线程会互相覆盖（实测 9 条记录被清成 0 条），后果是「关服自动安装已批准的版本」静默失效。现在写入是进程级加锁的重读-合并-写回，JSON 临时文件名带上进程与线程号。mods/留下半截新 jar，改成同目录临时文件 + 原子替换，残留会点名（MUC-10，P2）；台账里的文件名做安全校验、..不再被接受（MUC-18，P3）；下载也计入限速、失败文案不再谎报重试次数（MUC-21/29）。上游与检查
check.include_beta/include_alpha之前完全不起作用（MUC-02，P1）：频道过滤函数没有任何调用者，默认配置下 beta 会被当成可用更新推荐。现在生效；某个项目在所选频道没有版本时回落到更旧的频道并在报告里说明。sources.modrinth不再污染解析缓存（MUC-04，P1）：以前「没问过」被写成「确认不存在」，重新打开后这些 jar 在 24 小时内一次名字搜索都不会发。close()之后不再发请求（MUC-11）；每线程一个requestsSession（MUC-22）；超时拆成 connect/read 并给下载算总时长上限（MUC-27）；JSON 响应加上限（MUC-28）。健壮性
project-map.json存成 GBK / ANSI 不再让整次检查失败（MUC-03，P1）：以前UnicodeDecodeError会让!!muc status整页报错、检查以「check failed」收场。mods/读不到时不再报成「0 个 jar」：状态页写明原因，检查直接跳过（MUC-07 + 入口层接线）。scan_mods/classify_downloaded的异常收紧（MUC-07/08）；备份名判定改fullmatch（MUC-17）；加密或损坏的 jar 不再打破scan_jar的「不抛异常」承诺（MUC-19）；元数据解压加 4 MiB 上限（MUC-20）。入口层
!!muc status改成两段式（MUC-15，P2）：命令线程先回一句「正在读取」，扫描与渲染在后台线程完成后整屏给出——不再占用 MCDR 唯一的同步执行器线程；同一时间只允许一份状态页在生成。except补日志，_server_running失败时沿用上次已知状态（MUC-13，P2）：以前探测异常会静默丢弃所有玩家通知。!!muc别名下不再写死!!modupdate（MUC-23）。工程与工具
tools/mutation_check.py：把每条守卫逐个改坏（在仓库的临时副本里改，工作树不受影响），对应的用例必须失败；「没被抓住」与「锚点挪走」分开报告——这取代了 CHANGELOG 里「27 处注入验证」那种不可复现的说法，并修掉tests/test_docs.py指向两个从未进过仓库的脚本的悬空引用。tools/check_python_floor.py与 CI 的python-floor作业：每个文件都必须能以 3.8 解析、且不用更新的注解写法或标准库名字（ruff.toml的target-version只是配置，这一条才是检查）。bench/的 7 个测量脚本连同bench/README.md进仓库，并新增tests/test_documented_paths.py：文档里写出的仓库路径必须存在（此前README-dev.md引用了 7 个从未提交过的脚本）。lint/python-floor/mutation作业，两个 workflow 补permissions/timeout-minutes/concurrency；打包器包内 changelog 只留最新一节（已发布的 v1.6.0 里那份是 28,363 B、占产物 29.4%，且与自身页眉矛盾）。tests/test_docs.py的SECTIONS补上cleanup（此前整个cleanup.*命名空间不被校验；补上之后当场抓到README-dev.md把函数名写成了配置路径的样子）。如何验证
pytest（含 2 项真实 MCDR 端到端）python tools/mutation_check.pytools/mcdr_matrix.py（2.13.0 / 2.14.1 / 2.15.0 / 2.15.7 / 2.16.0）python tools/check_artifact.pyv1.6.1且与元数据一致ruff check .= 0 项；tools/check_python_floor.pyModUpdateChecker-v1.6.1.mcdr91,947 B，sha25624f2f11e…（与自检一致）未跑：CI 本身（推送后由 GitHub 跑）。
lint/python-floor/mutation三个作业是新加的，若变红请把日志贴给我。其它
CHANGELOG.md换成 v1.6.1 一节；README 的配置表补上每个数值选项的钳制范围，以及「mods 目录读不到会说明原因」的注意事项。Content-Length预检与累计上限，但project_versions的默认过滤没有加——那会抹掉「无适配构建」与「没有适配当前游戏版本的构建」两种判定的区别，取舍写在review-batch2/fixes/MUC-core-fixes.md§⑥。