docs: reclassify conhost-fallback finding - in-box ConPTY slowness, not an engine deadlock
Discriminating experiments: legacy shell forced to in-box ConPTY (openconsole=off) runs the 50MB payload in 38-94s (0.5-1.3 MB/s, huge variance); the GPUI lab without the sideloaded pair completes in 84s. Both shells degrade identically, so the earlier stall diagnosis is corrected: no engine deadlock, nothing to fix engine-side. Deploying the OpenConsole/conpty.dll pair next to any nebula-family exe stays a hard ops rule.
This commit is contained in:
1 parent
0c8fd3bc79
commit
7727bcb779
1 file changed
+5
-5
+5
-5
@@ -52,10 +52,11 @@
|
||||
**G1 阈值定为:等网格下吞吐不低于旧壳的 90%——当前通过**。后续在等
|
||||
网格条件下复测一次收严。
|
||||
|
||||
测量中抓到的真 bug(引擎级):无侧载 OpenConsole/conpty.dll 对时走
|
||||
系统 conhost 回退,大流量下会话**完全停摆**(PTY 缓冲写满、相关进程
|
||||
全部 0% CPU)。旧壳发行版始终带侧载对所以从未暴露。修复列入 P2;
|
||||
运维规则:任何 nebula 系 exe 旁必须部署侧载对。
|
||||
附带判明(判别实验 2026-08-13,两壳分别以 `openconsole=off`/缺侧载对
|
||||
复测):无侧载 OpenConsole/conpty.dll 时走 in-box ConPTY(conhost),
|
||||
两壳同样慢 3~8 倍且抖动巨大(0.5~1.3 MB/s、单轮 38~94s、尾部可达
|
||||
分钟级,观感近似停摆)。这是 in-box host 自身行为,**不是引擎死锁**,
|
||||
引擎侧无从修复。运维硬规则:任何 nebula 系 exe 旁必须部署侧载对。
|
||||
- **1b IME 清单**:输出一页人工验收清单,等用户有空亲手敲。
|
||||
- **1c 视觉对账集**:`scripts/visual_parity.ps1`——固定样张(boxdraw 全家
|
||||
桶、块/浓度/象限、Powerline 分隔符、CJK 对齐标尺、256 色/真彩)注入
|
||||
@@ -76,7 +77,6 @@
|
||||
- config 共享化:还掉实验场里复制 config 解析的债——共享 crate,一份
|
||||
nebula.toml 两壳同读。
|
||||
- 会话/SSH 逻辑共享化评估(能抽则抽,抽不动的记录原因)。
|
||||
- conhost 回退路径的大流量停摆修复(P1-1a 抓到,见上)。
|
||||
|
||||
### P3 接入 `nebula_app`(GPUI 成为产品 UI 层)
|
||||
|
||||
|
||||
Reference in new issue
Block a user