今日重点：✍️ 写作——上一篇写作（10-08）拆的是封面信里的测试小节； 今天拆前面那段 patch 清单。别小看这几行： 维护者决定要不要花时间读你的补丁，靠的就是它。这篇 v6 封面把五个动作做全了—— 每一条都在回答同一个问题：我凭什么敢合它。
📖 原文（真实邮件，逐字引用）
/ section[2/3] drops two gpio entries from omap5-igep0050.dts that nothing reads: the binding and both drivers describe three gpios, and the extra two have no pinctrl group and no other consumer in the tree.
Patches 2 and 3 remove warnings that the bare conversion leaves behind: am5729-beagleboneai: 241 -> 237 dtbs_check warnings omap5-igep0050: 188 -> 186 dtbs_check warnings with no new warnings in either file.
出处：Chaosheng Qu，[PATCH v6 0/3] dt-bindings: display: ti, tpd12s015: convert, and fix what the conversion exposed， 邮件 Date 头 2026-10-12 20:09 北京（注意：radar 上的归档时间比它晚约 20 小时，那是归档戳，不是发信时间）。 这封是封面信本身——练习题的 source 链接直接指向它。
✍️ 动作一：一条 patch 一句话，并回答「为什么安全」
/ column★ … that nothing reads: the binding and both drivers describe three gpios …
原句 [2/3] drops two gpio entries from omap5-igep0050.dts that nothing reads: the binding and both drivers describe three gpios, and the extra two have no pinctrl group and no other consumer in the tree.
拆解 删代码的补丁最容易被质疑，所以作者用冒号把结构摆得很清楚：冒号前一句话说清「删了什么」，冒号后立刻给「为什么删了没事」——三条独立证据：binding 和两个驱动都只描述三个 gpio（说明多出来的两个不在契约里）、没有 pinctrl group、树里没有别的消费者。这就是内核写作的默认格式：声明 + 冒号 + 证据。
怎么用 写删除类 patch 时，别只说 removed X。套这个格式：drops X that nothing reads: <证据一>, <证据二>, <证据三>。nothing reads 是个很地道的说法——不是「没人用」，而是「没有任何代码读它」。
✍️ 动作二：把「修好了」变成可核对的数字
/ column★ 241 -> 237 dtbs_check warnings … with no new warnings in either file
原句 Patches 2 and 3 remove warnings that the bare conversion leaves behind: am5729-beagleboneai: 241 -> 237 dtbs_check warnings omap5-igep0050: 188 -> 186 dtbs_check warnings with no new warnings in either file.
拆解 「修复有效」是主张，不是证据。241 -> 237、188 -> 186 才是证据：维护者可以自己跑一遍复现。更关键的是最后那句兜底——with no new warnings in either file：光说「少了几条」不够，还得说明「没多出新的」。报收益必报副作用，这是内核写作的硬习惯。
怎么用 凡是能产生数字的改动（警告数、测试数、耗时、内存），都给 A -> B。再加一句兜底：「no new X」。
✍️ 动作三：带着别人的 Acked-by 前进，要交代你改了什么
/ column★ The only change 1/3 saw between his ack and v5 is …, and that does not change which device trees the binding accepts.
原句 The only change 1/3 saw between his ack and v5 is the ports $ref Rob asked for, and that does not change which device trees the binding accepts.
拆解 v5 不小心把 Conor 的 Acked-by 弄丢了，v6 捡回来。问题来了：捡回一个「对旧版本给的 ack」时，如果中间改过东西，这个 ack 还算数吗？作者主动把这件事摊开讲：his ack 之后只动了一处（Rob 要求的 ports $ref），而且那处改动不影响这份 binding 接受哪些设备树——言下之意：ack 依然有效。自己不交代，维护者就得自己去 diff 两个版本；交代了，对方一句话就能放行。
怎么用 复用别人的 tag 时，主动写清两件事：改了什么（the only change … is …）、为什么不影响该 tag 的有效性（and that does not change …）。
✍️ 动作四：主动披露撞车，把选择权交出去
/ column★ If his series is the one to carry this, say so and I will drop mine.
原句 Eduard Bostina posted the same conversion on 2026-07-30, <20260730092245.52034-1-egbostina@gmail.com>. If his series is the one to carry this, say so and I will drop mine.
拆解 作者自己发现有人三个月前做过同一个转换，于是主动写进封面：先给出对方那封的 Message-Id（可核验），再一句把决定权交出去——say so and I will drop mine。不写，等维护者自己发现，那就变成「你在隐瞒重复工作」；写了，就是「我在替你省事」。这是内核列表上最受尊重的一种做法。
怎么用 模板：X posted the same on , . If his series is the one to carry this, say so and I will drop mine.
✍️ 动作五：反驳自动审查的误报，要指出它为什么不成立
/ column★ The sashiko finding that gpios lacks "minItems: 3" is not valid.
原句 The sashiko finding that gpios lacks "minItems: 3" is not valid. Rob explained that dt-schema makes the two bounds equal when only one of them is given, so "maxItems: 3" already means exactly three.
拆解 自动审查机器人（Sashiko）提了一条 finding，作者认为不成立。注意他驳回的方式：不是「机器人不懂」，而是给出机制层面的理由——dt-schema 在只给一个边界时会把上下界设成相等，所以 maxItems: 3 已经意味着「恰好三个」。而且他注明这个解释来自 Rob（该领域维护者），等于给自己的判断加了背书。
怎么用 驳回任何审查意见（人或机器）都用这个结构：The finding that X is not valid. <来源> explained that <机制原因>. —— 先说结论，再给机制，最后给依据来源。
✨ 辅助彩蛋：Acked-by 到底承诺了什么
/ column★ Acked-by / Reviewed-by / Tested-by 的分量不一样
官方定义（节选） Acked-by: is not as formal as Signed-off-by:. It is a record that the acker has at least reviewed the patch and has indicated acceptance. / Acked-by: is also less formal than Reviewed-by:.
中文 Acked-by 没有 Signed-off-by 那么正式，它只记录「此人至少看过、并表示接受」；它也比 Reviewed-by 轻。三者分量大致是：Signed-off-by（我参与了这个补丁，愿意署名担责）> Reviewed-by（我按 Reviewer's Statement 做过技术审查）> Acked-by（我认可它落地，但不代表我做过同等深度的审查）。
为什么值得知道 因为读封面信时，tag 会告诉你「谁在为什么背书」。看到 Acked-by，别当成 Reviewed-by 读——投稿人自己也得清楚：攒到一个 Acked-by，不等于通过了技术审查。
出处 Documentation/process/submitting-patches.rst（Linux 内核树内文档，链接见下方「练习」节后的引用）。
✍️ 今日练习
/ column✍️ 你做了一个 dt-bindings 转换，转换过程暴露出两个既有问题。写三段英文：① 用一句话说清第二个 patch 干什么，并用「声明 + 冒号 + 证据」的格式给出「为什么它安全」；② 用数字证明修复有效，并兜底说明没引入新警告；③ 你带着一位维护者的 Acked-by 发新版，但 ack 之后你又改了一行——怎么交代才不让人怀疑这个 ack 失效了？
💡 显示答案
参考作答（仿写示范，非原文）：① [2/3] drops the two unused gpio entries from <file> that nothing reads: the binding and both drivers describe three gpios, and the extra two have no pinctrl group and no other consumer in the tree. ② Patches 2 and 3 remove warnings that the bare conversion leaves behind: <board-a>: 241 -> 237 dtbs_check warnings <board-b>: 188 -> 186 dtbs_check warnings with no new warnings in either file. ③ Carry <maintainer>'s Acked-by on 1/3. The only change 1/3 saw between his ack and this version is the $ref he asked for, and that does not change which device trees the binding accepts.
📖 原文：[2/3] drops two gpio entries from omap5-igep0050.dts that nothing reads: the binding and both drivers describe three gpios, and the extra two have no pinctrl group and no other consumer in the tree. / Patches 2 and 3 remove warnings that the bare conversion leaves behind: am5729-beagleboneai: 241 -> 237 dtbs_check warnings omap5-igep0050: 188 -> 186 dtbs_check warnings with no new warnings in either file. / The only change 1/3 saw between his ack and v5 is the ports $ref Rob asked for, and that does not change which device trees the binding accepts.
📎 原文引用 彩蛋引用的内核官方文档：Documentation/process/submitting-patches.rst（git.kernel.org，已实测可访问）。
每日一句：A cover letter is not a summary of your patches. It is the case for why they are safe to take.
仿写示范