Skip to content

the verify script's pty helper is not called pty.py, and the sandbox result #40

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Sunrisepeak merged 2 commits into from
Sep 20, 2026
Merged

the verify script's pty helper is not called pty.py, and the sandbox result #40

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f7b87fb
the verify script's pty helper is not called pty.py, and the sandbox …
Sunrisepeak Sep 20, 2026
1288ffe
the mirror defect of the previous wave is repaired, not only recorded
Sunrisepeak Sep 20, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
36 changes: 33 additions & 3 deletions .agents/docs/2026-09-20-term-pass-control-record.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,6 +97,34 @@ done

这一项是用本地工作树（path 依赖）测的，因为它跑在 index 注册之前；注册之后的同一条链由 §3.5 的沙箱脚本按发布版本重新验证一次。

### 3.5 沙箱：只写版本号，从已发布的 index 解析

`xlings subos use term014 --sandbox --cmd "… bash 2026-09-20-term-pass-control-verify.sh"`，CN 镜像，mcpp 2026.9.18.3：

```
== A. identity and mirror ==
ok: mcpp 2026.9.18.3
ok: xlings mirror is CN

== B. openkal 0.14.0 resolves and states its version ==
ok: header 0.14.0 implementation 0.14.0 pass_control 4

== C. the position is reachable as a macro and as a module constant ==
ok: kal::terminal::pass_control and KAL_TERM_PASS_CONTROL_M are the same position

== D. the interrupt keystroke arrives as data above openkal-musl 0.16.0 ==
ok: the transcripts agree, keystroke for keystroke

0 assertion(s) failed
```

B 这一行是版本修正的可观察形式：实现自述 0.14.0，而在本波之前每个实现都答 0.11.0。

两件在这一步才暴露的事，都已修：

1. **index 是按构件分发的，不是按 main 分支。** 合并之后 `Publish Index Artifact` 工作流把 `pkgs/**` 发成 `xlings-res/mcpp-index@v<sha>`，客户端同步到的是那个构件；在它发完之前，沙箱解析 0.14.0 会报「not found in the synced index (mcpplibs@artifact:8d87b18)」。等构件发布并 `mcpp index update` 之后一次通过。
2. **脚本里的 pty 辅助文件不能叫 `pty.py`。** Python 把脚本自己所在目录放在导入路径最前面，于是它遮蔽了标准库的 `pty`，报 `AttributeError: partially initialized module 'pty' has no attribute 'fork'`——而脚本把这次失败归到了「本机 C 库也没收到这个按键」，即把自己的缺陷说成了对照组的。改名为 `ptykeys.py`。

## 4. 本波修掉的、与设计无关但同源的缺陷

- **`KAL_VERSION_MINOR` 停在 11**：整个生态的 `kal_version` 都答 0.11，条款 6.2 给 load 期绑定消费者的比较形同虚设。修正并加了检查。
Expand All @@ -119,9 +147,11 @@ done

mcpp-index PR [#442](https://github.com/mcpplibs/mcpp-index/pull/442) 一次注册以上九个描述符。

### 5.1 发布惯例上发现的一处缺陷（上一波留下的，本波未能修好）
### 5.1 发布惯例上发现的一处缺陷（上一波留下的，本波已修好）

`openkal-musl 0.15.0` 与 `openkal-llvm-runtime 0.11.0` 在 GitHub 上的标签带 `v` 前缀，而描述符的 GLOBAL 地址按惯例写无前缀形式，**这两个版本的 GLOBAL 地址曾返回 14 字节的 Not Found**。只补一个无前缀标签修不好：GitCode 上的资产不是 GitHub archive 本身（`75803192…` 对 `ee953bd8…`），而一个版本只有一个 sha256，所以无论地址怎么写两边都不可能同时对。

`openkal-musl 0.15.0` 与 `openkal-llvm-runtime 0.11.0` 在 GitHub 上的标签带 `v` 前缀，而描述符的 GLOBAL 地址按惯例写无前缀形式，**这两个版本的 GLOBAL 地址返回 14 字节的 Not Found**。补一个无前缀标签也修不好：GitCode 上的资产不是 GitHub archive 本身（`75803192…` 对 `ee953bd8…`），而一个版本只有一个 sha256。要修得把 GitCode 资产换成 GitHub 的 archive，而同名资产 `gtc` 拒绝覆盖。已在 index PR 里记录，留给仓库所有者。
**改源头而不是改描述符**：仓库所有者在 GitCode 侧删除了那两个资产之后，两个仓库各补了指向同一提交的无前缀标签，该标签的 GitHub archive 用 `gtc` 上传顶替原资产，描述符写这同一个文件的哈希。下载两边比对：`openkal-musl 0.15.0` = `ee953bd8…`、`openkal-llvm-runtime 0.11.0` = `b9b8eddb…`，GLOBAL 与 CN 逐字节相同。mcpp-index PR [#443](https://github.com/mcpplibs/mcpp-index/pull/443)。

本波九个包都按惯例发布：无前缀标签，GitCode 资产是 GitHub archive 原样上传，两边 sha256 相同（openkal 0.14.0 下载两边比对确认）。

Expand Down Expand Up @@ -151,4 +181,4 @@ mcpp-index PR [#442](https://github.com/mcpplibs/mcpp-index/pull/442) 一次注

这项缺陷早于本波，影响面小（只在没有尺寸的伪终端上），修法也小（`terminal.h` 一句 + 两个实现各两行），但要动三个包的版本，所以记在这里留给下一波，和「打开 macOS 的伪终端行」一起做。

**其余已知且记录在案的限制**：输出后处理（`OPOST`）不在模式字内，raw 模式下写 `\n` 仍会先得到回车；混合保留态恢复为环境惯常的集合；`VMIN`/`VTIME` 不可表达，要「会放弃的读」用 `kal_timeout_read`；openkal-musl 0.15.0 与 openkal-llvm-runtime 0.11.0 的 GLOBAL 镜像地址仍然取不到东西（上一波留下，修法需要所有者在 GitCode 侧删除同名资产）。
**其余已知且记录在案的限制**：输出后处理（`OPOST`）不在模式字内，raw 模式下写 `\n` 仍会先得到回车；混合保留态恢复为环境惯常的集合；`VMIN`/`VTIME` 不可表达，要「会放弃的读」用 `kal_timeout_read`；openkal-musl 0.15.0 与 openkal-llvm-runtime 0.11.0 的镜像不一致（上一波留下）已在本波修好，见 §5.1。
13 changes: 10 additions & 3 deletions .agents/docs/2026-09-20-term-pass-control-verify.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -202,7 +202,14 @@ if [ $rc -ne 0 ] || [ -z "$bin" ]; then
fail "the terminal probe did not build above openkal-musl $MUSL"
printf '%s\n' "$out" | tail -10
else
cat > "$work/pty.py" <<'PY'
# NOT NAMED pty.py. Python puts the script's own directory first on the import
# path, so a helper called pty.py shadows the standard module it imports:
#
# AttributeError: partially initialized module 'pty' has no attribute 'fork'
#
# Measured in the sandbox, where it turned D into "not run" and said the
# machine's own C library was at fault.
cat > "$work/ptykeys.py" <<'PY'
import os, pty, select, sys, time
marker, keys, cmd = sys.argv[1].encode(), bytes.fromhex(sys.argv[2]), sys.argv[3:]
pid, fd = pty.fork()
Expand All @@ -224,10 +231,10 @@ PY
if ! command -v python3 >/dev/null 2>&1; then
skip "D: no python3 here, so nothing can type at a pseudo-terminal"
else
port=$(python3 "$work/pty.py" reading 61620371 "$bin" 2>/dev/null)
port=$(python3 "$work/ptykeys.py" reading 61620371 "$bin" 2>/dev/null)
cc="$(command -v cc || command -v gcc || true)"
if [ -n "$cc" ] && "$cc" "$work/term/src/main.c" -o "$work/control" 2>/dev/null; then
ctrl=$(python3 "$work/pty.py" reading 61620371 "$work/control" 2>/dev/null)
ctrl=$(python3 "$work/ptykeys.py" reading 61620371 "$work/control" 2>/dev/null)
if printf '%s\n' "$ctrl" | grep -q 'byte 0x03'; then
if [ "$port" = "$ctrl" ]; then
ok "the transcripts agree, keystroke for keystroke"
Expand Down
Loading