Toby
Toby
I can confirm this. My project has seen a lot of users getting "sendmsg: invalid argument" caused by ECN. I guess we should have ECN detection, something similar to `isGSOSupported`?
@aggresss Thanks for your amazing work. **However I believe there's a serious bug when using this over localhost connections (at least on Windows):** **Steps to reproduce the problem:**. 1. Create...
Alloc & CPU pprof files of the server **(during connection hanging & CPU 100%)** are attached. ``` Showing nodes accounting for 7702.53MB, 99.90% of 7710.56MB total Dropped 20 nodes (cum...
Sorry but I don't see the point of this change. You are merely moving this function into a new struct with only one method. This is not abstraction - in...
> trojanUpUB 在 Trojan-killer 中目前是 [750](https://github.com/XTLS/Trojan-killer/blob/02c3683d1e1dba7a4f91462c7831f5caec0e15fc/main.go#L72)，而在 OpenGFW 中是 1000，应该会增加误报率，这是为了匹配内层 ECH 吗 是的，我测试的时候发现 750 会匹配不到浏览器发起的很多连接（curl 之类的倒是没问题），必须改到 1000 才行。原因就是 ECH 和一些乱七八糟的额外 extensions > 若从封锁的角度来说，我觉得应该持续监测同一目标（IP+端口+域名）的 M 条连接，若阳性率大于 N 就封锁该目标（用户可自定义），这样可以有效提升精准度 我也想过这个问题，只是目前单纯按 expr 表达式匹配的 ruleset...
> ~看起来 Trojan-killer 需要更新 2.0 了~，虽然增大匹配范围必定会增大单个连接的假阳率，但配合“封锁算法”应该仍能做到足够精准 更新的话要不直接在 OpenGFW 上 PR 吧 🗿
暂时不能，不过确实有计划
这种一般是 libc 对不上导致的，用 `CGO_ENABLED=0` 编译纯静态版本试试呢
Many thanks for bringing this up, but it's been fixed as part of #98 so I'm closing this
You can use IP instead of domain as server address