Skip to content

Commit 9fa3e53

Browse files
committed
plan: C4's 'does not reproduce' was measured on a different version
The record said C4 (`memset_pattern16`) was closed and did not reproduce. That measurement used a newer runtime. `lsp-mcpp-private`'s CI, pinned to `openkal-llvm-runtime@0.10.0`, still fails on it today: ld64.lld: error: undefined symbol: memset_pattern16 So the correct statement is "a runtime release fixed it", not "it does not reproduce". For a consumer still on the old pin those differ completely: the first gives an action, the second sends them looking for something intermittent that is not there. A defect's non-reproduction has to name the version it did not reproduce on. The same CI run surfaced something harder. That pin is mcpp 2026.9.17.3 and the index floor is 2026.9.18.3: error: index requires mcpp >= 2026.9.18.3 but this is mcpp 2026.9.17.3 [E0006] The index was closed to that client entirely, so none of its readings were about this wave --- they were about not being able to open the index.
1 parent d38b841 commit 9fa3e53

1 file changed

Lines changed: 27 additions & 1 deletion

File tree

‎.agents/docs/2026-09-21-openkal-ecosystem-completion-and-acceptance.md‎

Lines changed: 27 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -443,6 +443,32 @@ macOS,缺 SDK):
443443
444444
### 2.3 实现侧(openkal-*)
445445
446+
#### C4' — 「不复现」是在另一个版本上量的(2026-09-21 更正)
447+
448+
本文把 C4(`memset_pattern16`)记成「已通过,本机实测;关闭;不复现」。**那次测量用的是
449+
较新的 runtime**,而 `lsp-mcpp-private` 的 CI 钉在 `openkal-llvm-runtime@0.10.0` 上,它
450+
今天仍然红:
451+
452+
```
453+
ld64.lld: error: undefined symbol: memset_pattern16
454+
test_archive ... FAIL (compile, 0.83s)
455+
```
456+
457+
所以正确的陈述是**「被某个 runtime 版本修掉了」**,不是「不复现」。两者对一个还钉在旧版本
458+
的消费者来说结果完全不同——前者给出可执行的动作(升版本),后者让他去找一个不存在的
459+
偶发因素。**一个缺陷的「不复现」必须带上它是在哪个版本上不复现的。**
460+
461+
同一次 CI 还挖出一条更硬的:那个钉是 `mcpp 2026.9.17.3`,而索引的 `min_mcpp` 是
462+
`2026.9.18.3`——**钉在下限之下**:
463+
464+
```
465+
error: index requires mcpp >= 2026.9.18.3 but this is mcpp 2026.9.17.3 [E0006]
466+
```
467+
468+
整个索引对那个客户端是关着的。它的每一条读数都不是关于这一波的,而是关于「打不开索引」的。
469+
**一个落后于引擎的 CI 钉不是保守选择,是在测别的东西**;落到下限之下时,连那个「别的东西」
470+
也不是了。
471+
446472
#### I1 — P4:合成节点层
447473
448474
**状态**:未实现。设计 §5.4。
@@ -739,7 +765,7 @@ openkal-musl 0.18.0,而 0.18.0 当时还没进索引,消费者自己的 CI
739765
| C2 | cmp-module | 从分母移除,理由可追溯,**且声明可被证伪** | **已通过**:`[not-portable]` + `compat.py check`,红/绿两向实测 |
740766
| C2' | curl | 配方缺陷,不是不可移植 | **已定位未修**:两个目标两个不同真因,见 `docs/openkal-compat.md` |
741767
| C3 | `arc4random_buf` | 「应当有的符号」CI 断言 | 已通过 |
742-
| C4 | `aarch64-macos --profile release` | 本机实测 | 关闭;不复现 |
768+
| C4 | `aarch64-macos --profile release` | 本机实测 | **更正**:不是「不复现」,是被某个 runtime 版本修掉了——旧钉上仍在,见下 |
743769
| I1 | 写死的节点集合 | conformance 逐条断言 | 未实现(第五批) |
744770
| I2 | `presents = "windows"` 的 C 库 | 引擎改动数为 0 | 未实现(第五批) |
745771
| **A1** | 三目标 `mcpp build` | 全绿 | **待跑**(须用已发布钉,见 §8 第二条) |

0 commit comments

Comments
 (0)