docs(release): record convergence evidence, RC, and support matrix

WHY
- The code fixes existed locally without same-SHA release material; the G0..G12
  gate matrix, G-C VM evidence, RC artifact and explicit non-claims must be
  reviewable together.
- The release-candidate workflow packaged no consumers and did not verify its
  own checksum in the clean-install directory.

WHAT
- Record the current Linux/VM evidence at f075c96 and keep remote CI/G-C as
  UNVERIFIED; overall decision DO NOT RELEASE.
- Add G-C evidence (7/7 + positive CLI + 47-binary workspace run), RC report
  with gate matrix, clean-install and checksum result.
- Update STATUS, COMPATIBILITY, CHANGELOG, README, tests/README, mapping spec
  and releasing procedure for the new condition/span/power/Tool behavior.
- Harden release-candidate packaging: include tests/consumers, SOURCE_SHA,
  rustc -vV, verify the checksum before extraction, and run independent
  Analysis/Trace validation from the extracted archive.
- Record the local convergence note in NEXT_RELEASE_CI_CONVERGENCE.md.

TESTS
- Local fmt/clippy/workspace PASS; G-C VM PASS at f075c96.
- Linux RC: sha256 -c OK and clean-install smoke PASS.

CI_URL
- none: Gitee remote, no GitHub credentials. Windows/macOS and same-SHA G-C CI
  remain the blocking gates.

BREAKING_CHANGE
- none in this docs/workflow commit.

SECURITY_SCOPE
- strict remains fail-closed; no new capability or platform claim. G-C is
  local-VM evidence, not a CI-signed security URL.
This commit is contained in:
cflmy committed 2026-11-20 15:51:48 +08:00
1 parent f075c967f2
commit 8af72620f1
17 files changed
+1744 -126

No files matched your search

+14 -3
View File
@@ -168,6 +168,10 @@ jobs:
install -m 0755 target/release/marqdo "$stage/"
cp LICENSE-MIT LICENSE-APACHE CHANGELOG.md COMPATIBILITY.md README.md "$stage/"
cp -r spec examples "$stage/"
mkdir -p "$stage/tests"
cp -r tests/consumers "$stage/tests/"
printf 'source_sha=%s\nversion=%s\ntarget=%s\n' "$GITHUB_SHA" "$version" "${{ matrix.target }}" > "$stage/SOURCE_SHA"
rustc -vV > "$stage/rustc-vV.txt"
tar -C dist -czf "dist/marqdo-${version}-${{ matrix.target }}.tar.gz" "$(basename "$stage")"
cd dist
if command -v sha256sum >/dev/null 2>&1; then
@@ -181,12 +185,19 @@ jobs:
run: |
set -euo pipefail
tmp="$(mktemp -d)"
tar -C "$tmp" -xzf dist/marqdo-*.tar.gz
cd "$tmp"/marqdo-*
cp dist/marqdo-*.tar.gz* "$tmp/"
cd "$tmp"
sha256sum -c marqdo-*.tar.gz.sha256 2>/dev/null || shasum -a 256 -c marqdo-*.tar.gz.sha256
tar -xzf marqdo-*.tar.gz
cd marqdo-*
./marqdo --version
./marqdo --help >/dev/null
./marqdo check examples/multi-module/index.mq.md
./marqdo check examples/multi-module/index.mq.md --refs --emit analysis.json
./marqdo run examples/hello.mq.md
./marqdo run examples/multi-module/index.mq.md --trace-file trace.json
python3 tests/consumers/analysis_consumer.py analysis.json
python3 tests/consumers/trace_consumer.py analysis.json trace.json
python3 tests/consumers/validate_protocols.py analysis.json trace.json examples/tool-unknown-name/tool.json
- uses: actions/upload-artifact@v4
with:
name: marqdo-${{ matrix.target }}
+33
View File
@@ -31,6 +31,18 @@ currently `0.1.0`; no release tag has been cut. Dates use ISO 8601.
Linux security-gate evidence report (`docs/security-gate.md`).
- `check_never_spawns_a_foreign_kernel`: proves static `check` does not spawn a
declared Foreign provider, while `run` on the same document does.
- Power/control-flow corpus: 10 new AST goldens, 8 semantic goldens, 4
compile-negative fixtures and 2 runtime-negative fixtures covering
right-associative `**`, unary-minus precedence, parentheses, float negative
exponents, Strong+CodeSpan payloads, branch/loop conditions, inert prose and
code fences, and source spans inside one/two-backtick code spans.
- Independent Tool Protocol v1 consumers `sample-view`, `sample-knowledge` and
`sample-proof` (analysis/trace readers) with install, dispatch and remove
tests; none is a hard-coded CLI command.
- Locked path-dependency content-tamper test: a changed dependency is rejected
without rewriting `marqdo.lock` or the materialized `.marqdo/env`.
- G-C lane honours `CARGO_TARGET_DIR`, so a run from a clean `git archive` with
an external target directory uses the binary it just built.
### Changed
@@ -41,6 +53,12 @@ currently `0.1.0`; no release tag has been cut. Dates use ISO 8601.
reporting an opaque mismatch.
- Worker plugin fixtures use the platform dynamic-library name and compiler
flags.
- List-item condition paragraphs now accept only inert description; an
executable `Strong`/`Emph` there is a hard `markdown.invalid_construct`
error instead of being silently swallowed. Condition expression spans are
mapped from the literal bytes inside the code-span backticks.
- Linux strict `inspect --platform` remains a snapshot, not a promise. A run
that loses namespace/resources between probe and launch must fail closed.
### Fixed
@@ -48,6 +66,17 @@ currently `0.1.0`; no release tag has been cut. Dates use ISO 8601.
goldens (reproduced with a fresh `core.autocrlf=true` clone).
- macOS CI plugin failure caused by assuming a Linux `.so` and Linux-only
manifest target.
- macOS/Windows `E0433`: platform detection now uses `#[cfg]` functions
instead of `cfg!` around `linux::probe()`, so the Linux module is not
referenced on other targets.
- Linux strict probe→launch race: the CLI contract and the sandbox seam now
require a non-zero, diagnosed, fail-closed refusal when the real supervisor
launch fails after a successful probe.
- Parallel security tests shared process-id scratch paths and deleted each
other's project; every case now has a unique path and the helper is compiled
once.
- Condition parse diagnostics pointed at the backtick instead of the missing
operand; the payload map now excludes delimiters.
### Security
@@ -59,6 +88,10 @@ currently `0.1.0`; no release tag has been cut. Dates use ISO 8601.
sandbox starts. Strict L2 inheritance is future work.
- Capability brokers and worker processes remain cooperative
authorization/crash isolation, not an OS sandbox.
- The G-C lane was re-run at `f075c96` from a clean `git archive` in a
disposable Ubuntu 26.04 VM as ordinary user `dev`: 7/7 adversarial tests,
positive CLI contract and forged-marker rejection PASS. This is still a local
VM result, not a self-hosted GitHub security-lane URL.
## Baseline `0.1.0` (before this stabilization slice)
+43 -18
View File
@@ -1,7 +1,8 @@
# Compatibility and support matrix
Snapshot: workspace version **0.1.0**, no release tag cut. This file records
what was actually verified and separates it from what is only implemented.
Snapshot: workspace version **0.1.0**, code/security SHA **`f075c96`**, no
release tag cut. This file records what was actually verified and separates it
from what is only implemented.
## Public protocol versions
@@ -24,33 +25,48 @@ portable/tested locally but the native platform lane has not run here;
| Feature | Ubuntu Linux x86_64 | macOS arm64 | Windows x86_64 |
|---|---|---|---|
| Markdown profile / AST goldens | PASS | UNVERIFIED (CRLF fix) | UNVERIFIED (CRLF fix) |
| Rust interpreter / semantics | PASS | UNVERIFIED | UNVERIFIED |
| Markdown profile / AST goldens (160) | PASS | UNVERIFIED (cfg fix + CRLF fix) | UNVERIFIED (cfg fix + CRLF fix) |
| Rust interpreter / semantics (44) | PASS | UNVERIFIED | UNVERIFIED |
| `run` / `call` / `check` / `inspect` / `fmt` | PASS | UNVERIFIED | UNVERIFIED |
| lock v3 path/Git graph + sync | PASS | UNVERIFIED | UNVERIFIED |
| Analysis v2 / Trace v2 + consumers | PASS | UNVERIFIED | UNVERIFIED |
| Tool Protocol v1 | PASS | UNVERIFIED | UNVERIFIED |
| Native ABI (`plugin inspect`, trusted call) | PASS (C fixture) | UNVERIFIED (host-native fixture) | UNSUPPORTED in CI lane (no `cc`) |
| Worker ABI + `fs.read` bridge | PASS (Unix) | UNVERIFIED | UNSUPPORTED (tests are Unix-gated) |
| lock v3 path/Git graph + sync + tamper | PASS | UNVERIFIED | UNVERIFIED |
| Analysis v2 / Trace v2 + Python consumers | PASS | UNVERIFIED | UNVERIFIED |
| Tool Protocol v1 + 3 independent consumers | PASS | UNVERIFIED | UNVERIFIED |
| Native ABI (`plugin inspect`, trusted call) | PASS (C fixture) | UNVERIFIED (host-native fixture) | **UNSUPPORTED** |
| Worker ABI + `fs.read` bridge | PASS (Unix) | UNVERIFIED | **UNSUPPORTED** (Unix-gated) |
| Foreign Python/Node | PASS (system `python3`/`node`) | UNVERIFIED | UNVERIFIED |
| Foreign pinned environment binding | PASS | UNVERIFIED | UNVERIFIED |
| `strict` OS isolation | PASS when runtime probe passes (VM-verified); no CI URL yet | FAIL-CLOSED | FAIL-CLOSED |
| `strict` Linux L1 (pure Marqdo + Math) | PASS, probe-gated, 7/7 VM lane | **FAIL-CLOSED** | **FAIL-CLOSED** |
| `strict` L2 (Worker/Foreign/native) | **UNSUPPORTED** (flags refused) | **UNSUPPORTED** | **UNSUPPORTED** |
The Ubuntu column is from this development host (Ubuntu 26.04 container). It is
**not** a GitHub Actions result: this checkout has no GitHub remote or CI
The Ubuntu column is from this development host (Ubuntu 26.04 container) and a
disposable Ubuntu 26.04 QEMU/TCG VM for the security lane. It is **not** a
GitHub Actions result: this checkout has a Gitee remote and no GitHub
credentials. A fresh local clone with `core.autocrlf=true` was used to
reproduce the Windows checkout byte behavior.
## Strict L1 scope (ADR-0021)
- Enabled only when the runtime Bubblewrap probe succeeds. Probe failure or a
probe→launch race exits non-zero with a diagnostic and **never** falls back
to `project`.
- Inside the sandbox: user/mount/PID/net/UTS/IPC namespaces, no-new-privs,
empty capabilities, read-only `/project`, private `/proc`/`/dev`/`/tmp`,
rlimits and wall-clock/output supervision; the child verifies `PID 1`,
`NoNewPrivs`, `CapEff=0`, a non-initial uid map and no non-loopback network.
- Refused before launch: `--allow-read`, `--allow-write`, `--allow-foreign`,
`--allow-plugin`, `--plugin`, `--trust-native-plugin`, `--kernel`,
`--trace-file`.
- Strict L2 (letting Worker/Foreign/native plugin run inside the sandbox) is
not implemented and must not be inferred from L1 passing.
## Minimum environment
- Rust 1.82+ (verified with 1.100.0).
- Linux/macOS: a C compiler for Native ABI fixtures; Python/Node only for
Foreign kernels.
Foreign kernels; `git` for Git dependencies.
- `strict` additionally requires Bubblewrap and a kernel/environment that
permits user, mount, PID and network namespaces. It is enabled only when the
runtime probe passes; the current development container fails that probe, so
strict is refused there. See [`docs/security-gate.md`](docs/security-gate.md)
and ADR-0021.
permits user, mount, PID and network namespaces. See
[`docs/security-gate.md`](docs/security-gate.md) and ADR-0021.
## Explicitly unsupported in this candidate
@@ -60,4 +76,13 @@ reproduce the Windows checkout byte behavior.
- Automatic uv/npm/venv provisioning; only an explicitly provisioned, pinned
directory is verified (and must contain a regular-file executable).
- Registry/package publishing, JIT, Knowledge/Proof/View built-ins,
sessions, and general-purpose CAS.
sessions, and general-purpose CAS. The three `sample-*` tools are test
fixtures for the generic Tool protocol, not product features.
- Windows/macOS release archives from this host; the GitHub release job may
build macOS but has not run for the convergence SHA.
## Release decision
**DO NOT RELEASE.** G1/G2/G3 (native CI) and G9 (self-hosted security-lane CI
URL) have no real remote PASS for this SHA. See
[`docs/evidence/RELEASE_CANDIDATE_0.1.0_f075c96.md`](docs/evidence/RELEASE_CANDIDATE_0.1.0_f075c96.md).
+15 -11
View File
@@ -15,16 +15,19 @@ plugin ABI v2 are not claimed to work here.
> **Status: stabilization candidate, not released.** The language, lock v3,
> Analysis v2, Trace v2, Tool v1, Plugin ABI v1 (native + worker) and pinned
> Foreign providers are implemented and covered by tests on Linux. The
> Windows AST and macOS plugin CI failures are fixed in source, but the native
> Windows/macOS lanes have not been re-run from this checkout. **`strict` OS
> isolation is available on Linux only when the runtime Bubblewrap probe
> passes**; it was verified against the full escape/resource matrix in a
> disposable Ubuntu VM (ADR-0021). macOS/Windows fail closed, strict L2
> (Worker/Foreign) is not implemented, and the broker/worker are not a
> sandbox. See
> macOS/Windows `E0433` platform-cfg failure and the Linux strict
> probe→launch race are fixed in source at `f075c96`, but the native
> Windows/macOS and GitHub Linux lanes have not been re-run from this
> checkout. **`strict` OS isolation is available on Linux only when the
> runtime Bubblewrap probe passes**; it was verified against the full
> escape/resource matrix in a disposable Ubuntu 26.04 VM at `f075c96`
> (ADR-0021). macOS/Windows fail closed, strict L2 (Worker/Foreign) is not
> implemented, and the broker/worker are not a sandbox. The release decision
> is **DO NOT RELEASE** until the remote gates have same-SHA URLs. See
> [`docs/STATUS.md`](docs/STATUS.md), [`COMPATIBILITY.md`](COMPATIBILITY.md)
> and [`docs/security-gate.md`](docs/security-gate.md). The single next-phase
> plan is [`docs/STABLE_RELEASE_IMPLEMENTATION.md`](docs/STABLE_RELEASE_IMPLEMENTATION.md).
> and [`docs/security-gate.md`](docs/security-gate.md). The single current CI
> convergence plan is
> [`docs/NEXT_RELEASE_CI_CONVERGENCE.md`](docs/NEXT_RELEASE_CI_CONVERGENCE.md).
## Build
@@ -78,8 +81,9 @@ executes anything.
## What works (implemented and tested)
* **P0 contracts** — frozen Markdown profile, AST → program mapping, strict
expression grammar, evaluation semantics; 150 AST goldens and 48 negative
compile-error fixtures (`tests/`).
expression grammar, evaluation semantics; 160 AST goldens, 44 semantic
goldens, 52 compile-error fixtures and 12 runtime-error fixtures
(`tests/`).
* **P1 interpreter** — native Rust tree-walk interpreter: values, scopes,
functions, static parameter inference, assignments, branches, loops,
returns, recursion, structured errors, fuel/stack/output budgets.
+525
View File
@@ -0,0 +1,525 @@
# Marqdo CLI:CI 收敛、语言边界回归与稳定发行实施规范
> **文档性质:** 可直接入库的执行规范 / Release Convergence Plan。供未阅读任何历史聊天的维护者、贡献者与 DeepSeek Harness 使用。
>
> **仓库:** `https://github.com/cflmy/marqdo-cli`;开发者的本地 Git 远端可能是 Gitee,不应混淆远端提交、GitHub 镜像与真实 CI 状态。
>
> **事实基线:** 2026-11-20,GitHub `master` `c9f01aa8fb127a2dcc125103ef794832984774ad`。最新公开 CI:[run 44692692318](https://github.com/cflmy/marqdo-cli/actions/runs/44692692318)。工作区当前版本 `0.1.0`;检索时 GitHub Releases 为空。**这不是已发布稳定版。**
>
> **实施原则:** 以仓库的 `spec/` 和已接受 ADR 为语义权威,以真实 CI/本地运行日志为实现证据。此文档只定义剩余收尾工作和验收顺序,不重开已解决的架构议题,不授予自动发布权限。后续每次变更先更新 `docs/STATUS.md` 事实基线。若阅读时 HEAD 已变化,重新取得该 SHA 的 CI 证据,并在本规范记录差异,不能机械照搬旧故障。
> **执行更新(2026-11-20,本地收敛):** 远程 `master` 仍为 `c9f01aa8`,本机新增收敛提交 `d05f36f` → `f075c96`,但**未推送**(本检出只有 Gitee 远端且无 GitHub 凭据)。`f075c96` 上:Linux 本地 `fmt`/`clippy`/`cargo test --workspace --locked` PASS;专用 Ubuntu 26.04 VM 普通用户 `dev` 从 `git archive` 重跑 G-C lane **7/7 + 正向 CLI PASS**(见 `docs/evidence/2026-11-20-strict-gc-f075c96.md`);Linux RC 已打包并通过外部干净安装与 checksum 校验。Windows/macOS 与同 SHA 的 GitHub CI、G-C 自托管 CI URL 仍为 **UNVERIFIED**,因此总体决定仍是 **DO NOT RELEASE**。下一动作是把 `f075c96` 或仅补文档的后继提交推送到 GitHub 并重跑全平台 CI 与安全 lane。
---
## 0. 一页决策摘要
目前 **不是重新开发 Marqdo CLI** 的阶段。Rust 语言实现、Markdown profile、表达式解析、lock v3、Analysis v2、Trace v2、Tool v1、Foreign 固定环境、Linux strict L1 已有实现;Linux strict L1 曾在普通用户运行的专用 Ubuntu VM 完成 **7/7** 安全测试,但其证据对应提交 `b6fc182`,**尚无同一发布 SHA 的远程安全 CI URL**。
最新 `c9f01aa8` 的 GitHub CI 状态:
| CI job | 观察到的结果 | 具体证据 | 发布影响 |
|---|---|---|---|
| `fmt + clippy + test (Linux x86_64)` | **FAIL** | `strict_mode_enforces_isolation_or_refuses_closed`: Bubblewrap 启动返回 `Resource temporarily unavailable`,其余已执行的 17 项 CLI 集成测试通过 | **阻断 G-B** |
| `runtime contract (macos-latest)` | **FAIL** | `marqdo-sandbox/src/lib.rs:115` 对 `linux::probe()` 的非 Linux 引用导致 `E0433` | **阻断 macOS 运行契约** |
| `runtime contract (windows-latest)` | **FAIL** | 同样 `E0433` | **阻断 Windows 运行契约** |
| `schema instance validation + consumers` | **PASS** | 真实 Analysis/Trace 实例、Tool manifest 的 Schema 验证及消费者通过 | 公共协议该 job 有有效证据 |
| `release candidate artifact` | **SKIPPED** | 上游 CI 不通过;且该 job 仅按 tag/手动触发运行 | 尚无合格发行产物证据 |
| Linux strict G-C 独立安全 lane | **无同 SHA 的 CI 记录** | 本地 VM 7/7 的历史证据在 `docs/evidence/2026-11-20-strict-gc.md` | 若发行宣称 Linux strict L1,**必须补同 SHA 安全 CI** |
**不要将“CI 通过 1/4”解释为“功能完成 25%”。** 四个工作流 job 是离散的质量门禁,不是均匀的功能权重;代码已经较完整,但当前可发布判定为 **DO NOT RELEASE**。也不要猜测“几天内发版”;只有逐项 PASS 才能发布。
### 0.1 实施范围与拒绝范围
本轮必须:修复三平台 CI,补充 Markdown–表达式跨层金样与语义测试,证明核心现有包/协议/Tool/ABI/Foreign 的支持范围,补充独立扩展消费者,完成发行候选与安全证据,更新文档和版本。
本轮不应引入:新的顶层命令、第二套 Markdown 解析器、改写 `**` / `` `...` `` 的正式职责、知识编译器、Lean 证明器、View 网页、自动 uv/npm 生态管理、通用容器平台、JIT、公共包注册表。除非**现有承诺的稳定路径存在阻断级缺陷**,否则这些是独立项目或后续版本工作。
## 1. 项目架构及必须保持的边界
### 1.1 目标与计算路径
Marqdo 是独立 Rust 实现的 Markdown 原生语言,**不是**把 Markdown 翻译成 Python 运行。程序计算过程是:
```text
.mq.md UTF-8 source
│
▼
marqdo-markdown:固定 Comrak/CommonMark/GFM/Math profile(唯一 Markdown 解析)
│ Canonical MdNode + byte Span
▼
marqdo-frontend:Heading 作用域/导入/参数推断/Markdown AST→IR
│ 条件头、分支、循环、调用、引用、源码位置
▼
marqdo-ir + marqdo-expression:可执行 AST/强类型表达式语法
│
▼
marqdo-runtime:Rust 原生树遍历执行 + 数值/控制流
│
▼
marqdo-host / marqdo-sandbox:能力授权、预算、可选外部 Kernel/ABI
│
└── 通过 marqdo CLI 输出结果、诊断和受控 Trace
```
扩展架构保持正交:
```text
marqdo tool install <通用第三方 Tool bundle>
│
└── marqdo <工具名> ... 由外部进程承担,不在核心硬编码工具名单
公共消费协议:Analysis v2 / Trace v2(由 check/run 产生,Tool 可按需消费)
运行时扩展:Plugin ABI v1(独立于外部 CLI Tool)
Foreign:显式 Kernel Provider 与已固定环境(独立于 Tool 和 ABI)
```
**不准**为了测试方便把第三方工具、数学扩展或安全子系统直接塞进解释器的 `match` 链;**不准**将 Markdown 解析换成基于正则的近似语法;**不准**让 `check` 执行副作用。
### 1.2 规范、状态与历史的权威顺序
- `spec/markdown-profile.md`、`spec/mapping.md`、`spec/expression.ebnf`、`spec/evaluation.md`:语言语义。
- `spec/package-format.md`、`spec/analysis-v2.md`、`spec/trace-v2.md`、`spec/tool-v1.md`、`spec/plugin-abi.md`、`spec/foreign.md`、`spec/security.md`:工程契约。
- ADR-0016~0021:lock v3、Worker/Host、Foreign 环境绑定、Linux strict L1 的有效决策。
- `AGENTS.md`、`docs/STABLE_RELEASE_IMPLEMENTATION.md`、`docs/contributor-guide.md`:既有开发和稳定化规程;**本文细化其当前 CI 收尾工作,不推翻其范围约束**。
- `docs/STATUS.md`、`COMPATIBILITY.md`、`docs/security-gate.md`:必须逐 SHA 更新事实证据。旧状态、旧本地日志、旧 CI 不能自动转成新 SHA 的 PASS。
### 1.3 版本与安全概念
- **G-A**:核心运行与语义可用(开发阶段)。
- **G-B / Core Stable**:承诺的语言/工程功能在所有支持平台的 CI 通过;契约、工具与发行产物可复现。
- **G-C / Linux strict L1**:独立安全 lane 在普通用户、真实 OS namespace 环境中通过逃逸矩阵,并记录同一 SHA 的 CI URL。L1 仅含纯 Marqdo、console 和原生 Math;Foreign、Worker、Native ABI 等 **L2 仍不支持**,必须拒绝。
- 稳定版本号可以仍为预 1.0 版本;“stable”指受支持功能通过公布的验收门槛,**不等于**所有未来设想均已完成。
## 2. 当前失败的精确根因与首轮修复
### P0-1:Windows / macOS 的 `E0433` 条件编译错误
**已观察到:** 现有 `crates/marqdo-sandbox/src/lib.rs` 中的 `linux` 模块由 `#[cfg(target_os = "linux")]` 条件编译,但 `PlatformCapabilities::detect()` 使用:
```rust
let strict = if cfg!(target_os = "linux") {
match linux::probe() { /* ... */ }
} else { /* ... */ };
```
`cfg!(...)` 产生布尔常量,**不移除代码分支的类型检查**。macOS/Windows 仍需解析 `linux::probe()`,因而报 `E0433`。这是确定的源代码缺陷,而不是 CI 环境错误。
**最小正确修复:** 把平台专用操作分配给带 `#[cfg]` 的实现函数/模块:
```rust
#[cfg(target_os = "linux")]
fn detect_strict() -> StrictIsolation {
match linux::probe() {
Ok(mechanism) => StrictIsolation::Available { mechanism },
Err(reason) => StrictIsolation::Unavailable { reason },
}
}
#[cfg(not(target_os = "linux"))]
fn detect_strict() -> StrictIsolation {
StrictIsolation::Unavailable {
reason: format!("no verified strict OS adapter for {}", std::env::consts::OS),
}
}
```
然后 `PlatformCapabilities::detect()` 只调用 `detect_strict()`。这段是**示意修改**,开发者必须依仓库现有 API 实际实现,并消除非 Linux 专属的 `libc` 依赖警告。不要将 `mod linux;` 变成无条件导入,那会使 Linux 特有依赖污染其他平台。
**必须添加的验证:**
1. Ubuntu 本地 `cargo fmt --all -- --check && cargo clippy --workspace --all-targets -- -D warnings && cargo test --workspace`。
2. Windows/macOS **真实本机 CI** `cargo test --workspace`,不能仅 `cargo check`。
3. 在非 Linux 上 `marqdo inspect --platform --json` 报 `strict=unavailable`;`marqdo run ... --mode strict` 必须拒绝,未产生程序副作用。
4. 再次运行 plugin host-target fixture;确认先前的 `.dylib`/CRLF 问题没有在恢复编译后重新显现。
**完成定义:** 两平台的 `E0433` 消失;各自完整测试 job 绿灯;严格模式仍 fail-closed;没有更改公开语义。
### P0-2:Linux CI 的 strict probe 与运行时 `EAGAIN`
**已观察到:** `crates/cli/tests/integration_cli.rs::strict_mode_enforces_isolation_or_refuses_closed` 首先通过 `inspect --platform` 判断 strict 是否 available;如果 available,它就要求另一进程 `run --mode strict` 必须退出 0。最新 CI stderr:
```text
bwrap: Creating new namespace failed: Resource temporarily unavailable
```
因此日志支持的是**运行时受到瞬时或配额类环境限制**,尚不能证明具体是 `user.max_user_namespaces`、用户进程限制、宿主工作负载、Runner 策略中的哪一项;更不能断言 Linux strict L1 安全边界失效。
当前 `crates/marqdo-sandbox/src/linux.rs` `probe()` 会运行 Bubblewrap 烟测;`run()` 内部又调用 `probe()`,之后才真正启动带资源限制和项目挂载的监督进程。探测成功不代表后续启动一定成功,两次启动之间的系统资源与权限可改变;这属于典型 TOCTOU / 资源可用性问题。
**不允许的“修复”:** 删除测试、在 strict 失败时降级为 project、把 `EAGAIN` 误标为成功执行、在普通 CI 下断言所有 Linux 主机支持 strict、用 sudo/root 运行测试代替普通用户验证。
**推荐分离两项不同契约:**
1. **普通 CI:安全失败契约。** 在任意 CI 机器上都能运行。若本次实际隔离启动成功,则程序输出必须正确;若由于真实系统资源或权限不足启动失败,则必须退出非零、给出明确诊断、**项目代码从未在不受限环境执行**,不得隐式降级。`inspect` 仅是当时环境快照,不能拿它当“今后一定成功”的永久承诺。
2. **专用 G-C 设备:正向安全契约。** `MARQDO_REQUIRE_STRICT_SECURITY=1`、普通用户、可重建的 Ubuntu VM/self-hosted runner。此处 **不允许 skip/允许拒绝**;必须成功完成 run/call 和全部 7 项对抗测试,否则 G-C FAIL。
**实现建议:**
- 在受控测试中注入可预测的 `probe succeeds → launch returns EAGAIN` 情况(通过隔离的 adapter abstraction/test seam,而非改写全局 PATH 造成工具替换风险),断言 `LaunchFailed/StrictUnavailable` 被诊断且没有无沙箱执行。
- 取消普通 CI 的“第一次 probe 为 available,第二次执行就必须成功”的不可靠断言;仍要验证成功路径和失败路径,而不是简单忽略失败。
- 如果要减少多次 probe,只在确保**实际启动**使用同等/更强参数且执行时失败仍 fail-closed 的前提下实施;不要用纯文件存在或 bwrap `--version` 代替真实 probe。
- 记录 `EAGAIN` 时使用的 uid、namespace/sysctl、RLIMIT、bwrap 版本和进程负载作为诊断信息;不得将这类失败归结为不存在的语义错误。
- bounded retry 只能作为可选可靠性工程,不能吞没最终错误、不能绕过安全约束;默认先保证 fail-closed。
- 将错误码和机器可读错误分类稳定化(安全不可用、启动失败、超时/资源耗尽),不要把下游 bwrap stderr 当作任意调用成功。
**完成定义:** 普通 Linux CI 与受控 G-C 安全 lane 同时成立;在宿主不可用时 100% fail-closed;在专用 VM 中 7/7 且 run/call 正常。保存当前目标 SHA 的两条 URL。
### P0-3:修复后必须重新确认 CI,不得“预判绿灯”
当前 macOS/Windows 在编译阶段中止,所以不能据此声称以前的 native plugin、AST CRLF、Tool、Foreign 以及跨模块测试全部通过。恢复编译后有可能暴露后续错误。
具体动作:
- 从同一 SHA 打开完整 CI run,逐 job 导出失败步骤、`thread ... panicked`、错误分类和目标平台。
- 不要只看 GitHub Workflow 总结;对每个失败 job 下载原始日志。
- 不论上一条失败多简单,修复后必须以**新的 HEAD CI** 为依据重新评估所有平台。
- 不对 macOS/Windows 的 Unix-only ABI/Worker 测试作“通过”宣称;它们若没有运行,应在 `COMPATIBILITY.md` 明确 UNSUPPORTED,而不是掩饰为 PASS。
## 3. Markdown 与乘方:从疑虑收缩为正式回归测试
### 3.1 已确定的语义,不得擅自改变
Markdown 源码经过冻结 Comrak profile。`##` 定义函数;普通函数段落中的 `Strong` 是可执行语句;`Emph` 是返回表达式;有序列表是分支,无序列表是循环;列表项第一段的**首个 CodeSpan** 是条件表达式;后续普通文字可解释目的;BlockQuote 是显式调用。完整计算表达式由独立 Marqdo 表达式解析器解释。
**用户本地实测**已经说明:适当空格使 Markdown 按预期生成 AST,乘方可以正确运行。故这一轮**不应以“修复乘方”为由重新设计强/斜体语法**。Markdown 标记本身具有分隔符词法规则:紧凑写法造成不同 AST 属于需要准确文档化的语言边界,而不是一律归咎解释器。
```markdown
## main
**n = 2**
**`n = n ** 2`**
> print(n)
```
以上是建议的、对 `**` 乘方运算最稳妥的可读写法:外围 Strong 定义执行语句,内部 CodeSpan 避免 Markdown 把计算运算符看作强调分隔符。对于本机已经验证能形成正确 AST 的空格写法,也应以 AST 金样记录,不应禁止。
**正式判断标准:** `marqdo inspect --ast` 的 Comrak AST 是否对应规范结构;其后 `check`/`run` 是否按该结构执行。不得为“猜测用户想写什么”而增加原始 Markdown 正则修复器,也不得因源码像表达式就跳过 Markdown AST。
### 3.2 乘方在各层的当前证据
- `marqdo-expression/src/lexer.rs`:`TokenKind::StarStar`。
- `marqdo-expression/src/parser.rs`:`parse_power()`;测试 `power_is_right_associative_and_binds_tighter_than_unary_minus`。
- `marqdo-runtime/src/value_ops.rs`:`BinaryOp::Pow`;整数精确幂、负整数指数类型错误、指数最大 100000、估算结果位数上限 1000000;浮点由 `powf()` 处理。
- Runtime 层测试 `text_concatenation_and_power` 覆盖 `2 ** 10 = 1024` 与负指数拒绝。
- 当前 36 组 `tests/conformance/semantics/` 的 `.mq.md` 金样未单独测试 **Marqdo 原生** `**`;`36-foreign-call-python.mq.md` 的 `r ** 2` 属于 Python Foreign,不等价于 Marqdo 乘方。
### 3.3 新增的正向语义 fixtures(至少)
遵循 `crates/cli/tests/conformance_semantics.rs` 已有的 `.mq.md` + `.expected.json` 机制。不要用 `UPDATE_SEMANTICS_GOLDENS=1` 把错误期望直接覆盖;先手工计算期望。
| 编号建议 | 源码表达 | 预期 | 证明什么 |
|---|---|---|---|
| 37 | `*` + CodeSpan(`2 ** 3 ** 2`) + `*` | `Int 512` | Markdown→右结合解析→求值 |
| 38 | `*` + CodeSpan(`-2 ** 2`) + `*` | `Int -4` | 一元负号优先级 |
| 39 | `*` + CodeSpan(`(-2) ** 2`) + `*` | `Int 4` | 括号与优先级 |
| 40 | `*` + CodeSpan(`2.0 ** -1`) + `*` | `Float 0.5` | 浮点负指数 |
| 41 | `**` + CodeSpan(`n = 2 ** 3`) + `**`,然后返回 `n` | `Int 8` | Strong + CodeSpan 嵌套路径 |
| 42 | 有序列表条件 `` `n ** 2 == 4` `` + 叙述 + 引用调用 | 只调用真分支 | 条件 Markdown / 段落 / 控制流 |
| 43 | 无序列表条件 `` `n ** 2 < limit` `` + 叙述 + 递增 | 预期迭代和返回 | 循环、条件重新求值、乘方 |
例如,给 42 建立下述源文件;相应 expected 的 `stdout` 应为 `乘方正确\n`,返回值应为 `Unit`,实际 JSON 具体类型按已存在 wire-v1 金样约定构造:
```markdown
## main
**n = 2**
1. `n ** 2 == 4`,正确分支的自然语言说明。
> print("乘方正确")
2. `n ** 2 != 4`,另一分支的说明。
> print("乘方错误")
```
对 41:
```markdown
## main
**`n = 2 ** 3`**
*n*
```
### 3.4 必须新增的负向/边界测试
| 输入或场景 | 必须验证的行为 | 测试层 |
|---|---|---|
| `2 ** -1`,纯 Int | 类型错误;不是神奇地得到 `0.5` | Runtime error fixture |
| `2 ** 100001` | `resource.exhausted`(当前指数限额) | Runtime error fixture |
| `2 **` | 解析器诊断非法表达式,正确源码 span | 负向 fixture |
| 列表头 `` `n ** 2 ==` `` | 条件表达式语法错误,定位到原 Markdown 源码中缺失操作数的位置 | Frontend / CLI error |
| 多反引号 CodeSpan 包含 `**` | 仍能正确映射完整表达式与字节 Span | AST + frontend |
| `**n=n**2**` 等缺少空格的 Markdown 边界 | 固定 Comrak 真实 AST,不推断用户意图,不自动更正 | AST golden + 文档例子 |
| `**n = n ** 2**`(有空格) | 在实际 profile 中确认 AST,再据此固定接受/拒绝结果 | AST + CLI |
| 普通叙述段落出现 `` `n ** 2` `` | 不执行计算、不被作为分支条件 | 结构/副作用性质测试 |
| 代码围栏中出现 `**` | 保持惰性,不会触发运算 | AST + inert fixture |
| 纯说明段落的修改 | 可执行 IR/实际结果不变 | property test |
| 强行嵌套 Markdown 强调导致非法 payload | 给出诊断,不静默换另一计算规则 | negative test |
**条件头被吞掉的可执行节点:** 当前 `map.rs::condition_of()` 使用 `&item.children[1..]` 形成分支体,首段里第一个 CodeSpan 后若出现 `Strong` / `Emph`,这些可执行标记可能被整体忽略。建议在原有“首个 CodeSpan + 说明文本”语法上增加一个严格检查:首段剩余结构仅允许惰性叙述(按已冻结 profile 判定),保留其原始 AST 与 SourceSpan;若出现其他可执行角色,报 `markdown.invalid_construct`,不能无声丢弃。先补复现 fixture,再制定必要的精确规则;**无需更换 CodeSpan 条件头的正式语法**。
**CodeSpan 源码位置:** `condition_of()` 当前使用 `PayloadSegments::single(&literal, condition_node.span)`,须回归验证反引号是否计入字节跨度;若诊断偏移,修正 span 映射而不是改 Markdown AST。要求单/双反引号、中文前缀、换行和多字节 UTF-8 的准确来源测试。
### 3.5 正确的测试组织与执行命令
```bash
cargo test -p marqdo-expression power_is_right_associative_and_binds_tighter_than_unary_minus
cargo test -p marqdo-runtime text_concatenation_and_power
cargo test -p marqdo-cli --test conformance_ast
cargo test -p marqdo-cli --test conformance_semantics
cargo test -p marqdo-cli --test negative
cargo test -p marqdo-cli --test properties
```
逐个检查 `*.expected.json` 的类型标签、stdout、退出码和诊断分类;新增 AST goldens 先独立检查真实树,再写语义期望,不允许一键刷新旧金样掩盖回归。CI 应直接运行上述测试(或被 `cargo test --workspace` 包含),并在单独的 job step 报告 fixture 覆盖,使 Markdown 语义边界成为可发现的门禁。
## 4. 第二轮:核心生态、协议与跨平台验证(不能以 Linux 绿灯替代)
### P1-1:lock v3、环境与可复现性
既有 `marqdo.lock v3`、传递包实例图、Git cache、`sync --locked` 和环境原子物化需要继续验证,**不需要重写包管理器**。
- `cargo test -p marqdo-cli --test package_env`;增加/核实传递依赖固定、多版本别名、路径/Git 内容篡改、锁文件缺失/过时、`run`/`check` 不下载不改锁、原子安装失败可回滚。
- Windows 与 macOS 的实际原生测试必须覆盖承诺支持的路径与 Git 功能;Unix-only 设计要作为 UNSUPPORTED 写清。
- 校验包内容哈希策略对无关构建目录与特殊文件的一致处理,不能因 checkout 位置变化造成无端哈希漂移。
### P1-2:Analysis v2 / Trace v2
- 运行 `cargo test -p marqdo-cli --test analysis_trace`,及 `tests/consumers/validate_protocols.py`、`analysis_consumer.py`、`trace_consumer.py` 的真实实例验证。
- 验证跨项目移动后的逻辑身份、资源定位、SourceSpan、跨模块追踪、Trace 与 Analysis 字节摘要绑定、错误/截断/过期证明;不得把缺少某字段的 JSON 视为 Schema 检查成功。
- 新增乘方/控制流 fixture 后,证明 Analysis 的语义投影一致,Trace 只记录实际命中分支;阅读工具不执行源码。
### P1-3:通用 Tool Protocol v1 的独立消费者
目前已实现未知名字 `benchmark` 的 Tool fixture,尚缺用于验证三类真实消费者的独立样例。新增**三个小型、彼此独立、可跨平台运行**的模拟 Tool:`sample-view`、`sample-knowledge`、`sample-proof`。名字仅是测试 fixture,不是 CLI 硬编码白名单,也不是在本仓库实现实际 View/Knowledge/Proof。
各消费者从公共 `Analysis v2` 或 `Trace v2` 中读取极小字段、产生确定输出;按 Tool v1 manifest 安装、调用和卸载。至少验证:
- 安装 `marqdo tool install <manifest>` 后 `marqdo <自定义名称>` 正确分发;Tool 不在 CLI 源码内特殊注册。
- 传递原样 argv/cwd/stdin/stdout/stderr 与子进程退出码;打断、卸载、恶意 bundle 路径与 whole-bundle digest 篡改严格拒绝。
- Analysis/Trace 兼容版本不匹配时在调用前给出清晰诊断。
- CI 使用临时 `MARQDO_TOOL_HOME`;不触碰个人工具目录,不借由用户 PATH 自动运行未注册程序。
### P1-4:ABI/Worker/Foreign 的真实支持矩阵
- Linux/macOS Native ABI:使用各自主机目标构建 C fixture;成功调用、多导出、ABI mismatch、损坏 bundle、Worker 超时/崩溃。
- Windows:如当前没有承诺 Windows Native/Worker,明确 UNSUPPORTED;不得通过“测试跳过”声称可用。仍应验证 Windows 的语言/Tool/包/分析契约。
- Host Broker:只宣称已有明确实现的 `fs.read` 等能力;其他 `fs.write`、`net.request` 等返回 `capability.unavailable` 不算正常功能支持。
- Foreign:验证真正启动的可执行文件来自已 pin 的环境,环境内容 digest 与版本漂移检查,非法 wire、stdout/stderr 限额、超时;未实现 uv/npm/venv 自动生成,应在支持矩阵写清,**不作为稳定基础 CLI 的必需特性**。
- `check` 不执行 Foreign、Plugin 或用户程序;恶意输入下也必须保持静态。
## 5. G-C 专用安全验证与准入
### 5.1 已完成的事实与不可推断部分
ADR-0021 已实现 Bubblewrap supervisor,Linux strict L1 在专用 Ubuntu 26.04 VM、普通用户 `dev` 下通过 7 个对抗测试,包括身份/权限、主机读写、网络、进程与资源。日志位于 `docs/evidence/2026-11-20-strict-gc.log`,对应执行的代码提交为 `b6fc182`。
**不能据此推断:** 任意 Ubuntu 都支持 strict;所有 Linux 版本均通过;macOS/Windows 已具备 strict;Worker/Foreign/Native 插件能在 strict 中运行;已经获得 G-C 远程 CI 认证。
### 5.2 同 SHA 的安全 CI
- `.github/workflows/security-linux.yml` 已设定 `[self-hosted, linux, marqdo-security]`,必须由**专用、可销毁 VM** 执行。
- runner 主机需实际支持非 root Bubblewrap user/mount/PID/net namespace、正确 AppArmor profile;不可通过全局关闭保护或 sudo 模拟非特权用户。
- `MARQDO_REQUIRE_STRICT_SECURITY=1`:runner 缺机制、跳过测试、7 项中任一失败均视为 FAIL。
- 记录完整 run URL、被测 SHA、VM image/kernel、bwrap 版本、uid、源码与 artifact 摘要、7/7 结果;只有同 SHA 证据可作为正式发行审核依据。
- 严格证明“源码从未在降级模式运行”;启动/挂载/namespace 失败时退出非零并且没有副作用。
- L2(Worker/Foreign)明确拒绝;不应为赶发版扩展 strict 范围,除非另立 ADR 与攻击测试。
### 5.3 可选发行范围决策
**方案 A(推荐):** 发行声称 Linux strict L1,前提是同 SHA 专用安全 CI PASS。该方案最符合当前 `STATUS.md` 的发布门槛。
**方案 B(仅当项目负责人明确改变发行范围):** 发布仅 G-B 的普通 CLI,正式关闭/不宣称 strict L1,文档、功能开关、支持矩阵和测试必须一致,确保用户不会误以为安全运行已认证。不能一边保留可用的 L1 声明一边省略 G-C 证据。
## 6. 发行产物与工作流纠偏
当前 `release-candidate` job 依赖 Linux、macOS、Windows 核心 CI;在前置失败时被 skip 合理。它只对 release tag 或 `workflow_dispatch` 执行;`docs/releasing.md` 要求源代码与 CI 的 SHA 一致、干净环境安装、checksum、spec/examples、许可证、兼容矩阵与回滚说明。
### 6.1 发版前的工作流检查
- 修复 P0-1、P0-2 并触发**同一 SHA** 的全平台 CI,所有 jobs 必须 PASS。
- 单独跑 G-C 真实 VM(若宣称 Linux strict L1);记录 URL 与环境。
- 校验 Release job 输出的 OS/arch/target 与 `rustc -vV` 一致,不可以把实际 host 二进制错误标注成 `aarch64-apple-darwin`。Windows 未配置发行包就不要宣称提供 Windows 发行资产。
- Windows/macOS/Linux 原生 smoke(当前支持的功能)必须发生在提取后的干净目录中,不能只在源代码树里 `cargo run`。
- `git describe --tags`、`Cargo.toml` 版本、二进制 `--version`、归档名、`CHANGELOG.md`、compatibility、Release Notes 保持一致;拒绝不同 SHA 的证据拼接。
- 发布前先在 RC 上执行 `sha256sum -c`(macOS 对应 `shasum -a 256`),再安装、运行、通过 Python consumers 验证 Analysis/Trace。
- 发版 tag/release 是不可逆的对外动作,需项目负责人显式批准;Harness 只可预备 RC 和证据,不能因完成文档而自行发 tag。
### 6.2 CI 环境可靠性要求
- 固定必要工具版本/lock,所有主机对同份 `.mq.md` 拥有一致字节行为(`.gitattributes` 与 CRLF 金样)。
- 影响平台的改动强制在受支持的平台编译和完整测试;仅 Ubuntu 本地通过不能合并发布。
- 对 GitHub-hosted runner 资源不稳定场景采取安全失败检测,不能把所有 `strict` 失败一概当可用;对应专用安全 lane 必须跑真成功路径。
- 避免把工具链安装、网络下载、外部 Python/Node 可用性放进与语言语义直接耦合的金样,从而避免无关环境波动把语言回归误判。
## 7. 严格的发布门禁清单(唯一签署表)
每项只允许 `PASS(证据URL/日志/SHA)`、`FAIL(原因)`、`UNVERIFIED(待执行)`、`EXPLICITLY UNSUPPORTED(范围说明)`。**核心必需项不允许标成 UNSUPPORTED;可选能力明确不发布时才可以。**
| 门禁 | 当前 c9f01aa8 | 证据/下一动作 |
|---|---|---|
| G0 同 SHA、版本与变更集明确 | PASS(基线记录) | SHA c9f01aa8;0.1.0;无 release |
| G1 Linux fmt/clippy/workspace/smoke | FAIL | strict CI false assumption/EAGAIN |
| G2 Windows workspace + portable smoke | FAIL | cfg! 下 linux::probe E0433 |
| G3 macOS workspace + portable smoke | FAIL | 同 E0433 |
| G4 公共 Schema + 消费者 | PASS(当前 job) | 当前 GitHub CI schema job |
| G5 Markdown AST/乘方新增语义回归 | UNVERIFIED / 待新增 | 第 3 章 test matrix |
| G6 全套负向/源位置/惰性属性测试 | UNVERIFIED(同 SHA 完整 job 未绿) | 现有 `negative`/`properties` 加新增条件头测试 |
| G7 lock v3 与包环境受支持平台 | UNVERIFIED(跨平台未完成) | `package_env`、Git / lock 反例 |
| G8 Tool、ABI、Worker、Foreign 受支持矩阵 | PARTIAL | 新消费者 fixture,macOS Native 实测;Windows 明确范围 |
| G9 Linux strict L1 安全真实测试 | VM PASS(旧 SHA),同 SHA CI UNVERIFIED | self-hosted security lane |
| G10 RC 构建/校验和/跨平台干净安装 | UNVERIFIED | release-candidate job 尚未执行 |
| G11 文档与版本/许可证/兼容矩阵 | PARTIAL | `STATUS.md` 落后最新 CI,需修订 |
| G12 负责人批准 tag + Release | NOT STARTED | 前面门禁 PASS 后审批 |
**当前总体发布状态:`DO NOT RELEASE`。** 不给虚假的百分比;以余下硬门禁的通过证据量化进度。建议每次提交后在 `STATUS.md` 维护上述表,至少填 commit SHA、各 CI URL 和失败测试。
### 7.1 关闭条件:最短可衡量路径
- **第一个工作包完成:** G1~G3 全绿(解决旧 CI 三处失败);不再存在 E0433 / 环境依赖的错误断言。
- **第二个工作包完成:** G5/G6 全绿(乘方、Strong/CodeSpan、列表叙述、负向边界、源 Span);语言语义不需重新设计。
- **第三个工作包完成:** G7/G8 全绿或清楚地标为非承诺子能力;独立 Tool 消费无硬编码。
- **第四个工作包完成:** G9 由专用 VM 自托管 CI 在同 SHA PASS,或经负责人明确决定仅发行 G-B 并真正关闭 G-C 声明。
- **第五个工作包完成:** G10/G11 PASS、Release Candidate 清洁安装,再经批准完成 G12。
## 8. 分阶段执行与 PR 编排(供 Harness 持续推进)
| PR/工作包 | 优先级 | 改动范围 | 先写失败测试 | 所需 CI | Done 定义 |
|---|---|---|---|---|---|
| `RC-01 cross-platform cfg` | P0 | `marqdo-sandbox` 平台检测与依赖 cfg | 非 Linux compile + strict refusal | Linux/Windows/macOS | 编译、测试、拒绝契约绿 |
| `RC-02 strict CI robustness` | P0 | `linux.rs`、`integration_cli.rs`、安全适配层 seam | probe-success/launch-EAGAIN | 普通 Linux + VM | 两种环境均有正确契约 |
| `RC-03 Markdown power fixtures` | P0/P1 | AST/conformance/negative/property/spec 示例 | 新乘方 goldens、错误 spans | 全平台 workspace tests | 结果与语法精确一致 |
| `RC-04 consumers & platform coverage` | P1 | Tool fixture、包/ABI/Foreign CI | 独立名称+恶意 bundle+ABI | Linux/macOS/Windows 支持矩阵 | 无硬编码,边界清楚 |
| `RC-05 G-C security CI` | P1(若宣称 L1 则必需) | self-hosted VM lane 与证据绑定 | 真逃逸负例,不允许 skip | 安全 runner | 同 SHA 7/7 PASS |
| `RC-06 artifact and release` | 发布门槛 | 版本、文档、工作流、checksums | 干净安装的 smoke | 全平台/安全/RC | 可签署发行记录 |
**PR 原则:** 每 PR 对应一个可以独立关闭的错误类,绝不把跨平台修复、控制流语义重定义、外部产品开发打包成大重构;在新 CI 上出现额外故障则新增最小复现并纳入同类工作包。每个 PR 的提交信息至少包含:`WHY / WHAT / TESTS / CI_URL / BREAKING_CHANGE / SECURITY_SCOPE`。
## 9. Ubuntu 开发工作流与 DeepSeek Harness 权限
日常执行以普通用户进行:
```bash
pwd
git status --short --branch
git remote -v
git rev-parse HEAD
git fetch --all --prune
rustc -vV
cargo fmt --all -- --check
cargo clippy --workspace --all-targets -- -D warnings
cargo test --workspace --locked
cargo build --release --locked -p marqdo-cli
```
**重要:** 本地 `origin` 可能是 Gitee,而 GitHub 才运行 CI。Push 到 Gitee 的输出只是 Gitee 更新证据,不能证明 GitHub 镜像/Actions 已经检出了同一 SHA。先对比 GitHub `master` SHA,再读取 Actions 结果,不能把两个远端混为一谈。
可在专用、可销毁 Ubuntu VM 中给予 Harness 有审计的 sudo,安装测试依赖与配置隔离环境;但**实际 G-C 安全用例必须以普通用户运行**。若当前机器不能创建 namespace,则开发普通语言/CI 工作包可继续,G-C 转到已验证 VM,不能通过关闭 AppArmor 全局规则来凑成功。
日常建议按以下顺序:
```bash
# 快速复现核心行为
cargo test -p marqdo-cli --test integration_cli
cargo test -p marqdo-cli --test conformance_ast
cargo test -p marqdo-cli --test conformance_semantics
cargo test -p marqdo-cli --test negative
cargo test -p marqdo-cli --test analysis_trace
cargo test -p marqdo-cli --test tool_protocol
cargo test -p marqdo-cli --test package_env
# 安全专用测试:仅在能实际提供 OS namespace 的 VM 中强制执行
MARQDO_REQUIRE_STRICT_SECURITY=1 ./tests/security/run-lane.sh
```
每修完一个问题应先本地跑最小用例,再跑 workspace,再等待远端支持平台 CI;只有被 CI 验证的 SHA 才能进入下一门禁。不要跳过失败测试,也不要把 `cargo check` 的通过当作 `cargo test` 的通过。
## 10. 发行验收报告模板
在仓库中维护 `docs/evidence/RELEASE_CANDIDATE_<version>_<shortsha>.md`(文件名在创建时替换),填入下面模板,并将 CI 链接明确绑定 SHA:
```text
Marqdo CLI release candidate
Source SHA:
Workspace version:
Intended release target/platforms:
Linux fmt/clippy/workspace/smoke URL / PASS?:
Windows native runtime tests URL / PASS?:
macOS native runtime tests URL / PASS?:
Schema + independent consumers URL / PASS?:
Markdown/power AST, semantic, negative fixture results:
Package/Tool/ABI/Foreign supported-matrix results:
Linux strict L1 claimed? (yes/no):
If yes, G-C self-hosted security URL + 7/7 evidence SHA:
Unsupported optional features:
Release archive names, targets, SHA-256:
Clean-install smoke (external directory) outcome:
Changelog / COMPATIBILITY / STATUS links and version:
Known limitations and rollback approach:
Gate summary G0..G12:
FINAL DECISION: RELEASE / DO NOT RELEASE
Approved by / time:
```
必须通过:`G1~G8`、`G10`、`G11`,以及如果承诺 strict L1 则 `G9`;`G12` 是最后的人工批准。报告附全部原始日志或 URL。不得将未运行的 Windows/Worker/Foreign 自动环境功能写成已通过。
## 11. 可直接发送给 DeepSeek Harness 的完整任务指令
> 下面是执行指令,不是已完成工作的声明。用户允许在专用 Ubuntu 开发 VM 中按需 sudo,不代表允许绕过独立安全验证或未经许可发布 tag。
```text
你现在负责 cflmy/marqdo-cli 的稳定版发版收尾。工作以
`docs/NEXT_RELEASE_CI_CONVERGENCE.md` 为唯一当前 CI 收敛计划;
语言规范以 spec/、已接受 ADR、AGENTS.md 为准。
先获取最新的本地 HEAD、Gitee 和 GitHub master SHA、最近一次 GitHub CI。
若仓库已有新提交,必须重新定位失败,不许盲目修复文档中基线 SHA 的旧故障。
目标不是扩写 CLI,而是在不改变冻结语言语义的前提下完成:
(1) Windows/macOS 的 Linux-only cfg 编译修复;
(2) Ubuntu 普通 CI 下 strict probe 与实际启动竞态/资源不足的 fail-closed
契约,并在独立安全 VM 保证严格隔离正向执行与 7/7 攻击矩阵;
(3) 增加完整 Markdown AST、乘方、分支/循环条件、源 Span、惰性与
错误边界的黄金测试和端到端语义测试,不擅自改 CodeSpan 条件语法;
(4) 验证 lock v3、Analysis/Trace、完全通用的 Tool、ABI/Worker/Foreign
的支持矩阵和独立消费者;
(5) 完成全平台 CI、同 SHA 的 G-C 证据、Release Candidate 干净安装、
checksums、COMPATIBILITY、STATUS、CHANGELOG 与发版审批材料。
开发纪律:
- 修复前先写最小失败复现,修复后运行其正反测试与 workspace tests。
- 不删除/跳过失败测试;不改金样来掩盖未知变化;不降级安全模式。
- 不写死官方 Tool 名称,不引入第二解释器,不扩大顶层命令数。
- 对于未承诺的 Windows ABI、strict L2、uv/npm 自动生成环境,
明确 UNSUPPORTED,不为了“全部完成”强行加入 CLI。
- 在专用 VM 内可为环境配置使用 sudo;G-C 执行者必须是普通用户。
- 每个 PR 提供源 SHA、问题根因、代码位置、运行测试、GitHub CI URL、
对平台和安全承诺的影响。任何 CI 失败均视为门禁未通过。
- 最新仓库若出现额外阻断,增补最小再现测试及 STATUS 事实,继续
将问题关闭到 release gate PASS;不要反复只提交开发计划。
- 版本 tag / 正式 Release 需要项目负责人明确审批,不可自动发布。
最终交付:G0..G12 门禁矩阵、完整运行日志/CI 链接、可信安全范围、
可安装和验证 checksum 的 RC 产物、完整变更说明、可复现发行流程。
如果任一必要门禁没有真实 PASS,明确给出 DO NOT RELEASE 和下一步动作。
```
---
## 12. 外部事实与代码审查索引
- 最新审查提交:`https://github.com/cflmy/marqdo-cli/commit/c9f01aa8fb127a2dcc125103ef794832984774ad`
- 最新 CI:`https://github.com/cflmy/marqdo-cli/actions/runs/44692692318`
- 工作区状态:`docs/STATUS.md`、`COMPATIBILITY.md`
- 安全机制与证据:`docs/adr/0021-linux-strict-isolation.md`、`docs/security-gate.md`、`docs/evidence/2026-11-20-strict-gc.md`
- CLI 测试:`crates/cli/tests/integration_cli.rs`、`conformance_ast.rs`、`conformance_semantics.rs`、`negative.rs`、`properties.rs`
- Markdown/表达式:`crates/marqdo-markdown/src/profile.rs`、`crates/marqdo-frontend/src/map.rs`、`crates/marqdo-expression/src/parser.rs`、`crates/marqdo-runtime/src/value_ops.rs`
- 发行流程:`.github/workflows/ci.yml`、`.github/workflows/security-linux.yml`、`docs/releasing.md`
**证据边界说明:** 这里的 GitHub CI 结论来自该次真实 Actions 运行与原始任务日志;本文件编写期间没有在作者环境重新运行 Cargo/OS 隔离测试,也没有修改 GitHub 仓库。新增代码与测试为实施要求,不应被读成已完成。后续维护者必须将所有 `UNVERIFIED` 转为含原始来源的结果,才能签署稳定版本。
+91 -81
View File
@@ -1,7 +1,8 @@
# Marqdo Implementation Status
Snapshot: **2026-11-20**, workspace version `0.1.0`, no release tag.
Authoritative task plan: [`STABLE_RELEASE_IMPLEMENTATION.md`](STABLE_RELEASE_IMPLEMENTATION.md).
Authoritative CI convergence plan:
[`NEXT_RELEASE_CI_CONVERGENCE.md`](NEXT_RELEASE_CI_CONVERGENCE.md).
Support matrix: [`../COMPATIBILITY.md`](../COMPATIBILITY.md).
This file records what was actually executed on the current tree. "Implemented"
@@ -10,130 +11,133 @@ commands were run and where. The historical M0.5 snapshot is preserved at
[`history/STATUS-M0.5-snapshot.md`](history/STATUS-M0.5-snapshot.md) and must
not be read as current status.
## 1. CI baseline and current evidence
## 1. Remote baseline and current evidence
The last known public GitHub CI at baseline `e966b76` (run
[44680378977](https://github.com/cflmy/marqdo-cli/actions/runs/44680378977))
was **not green**:
GitHub `master` is still `c9f01aa8fb127a2dcc125103ef794832984774ad`; the latest
public CI is run
[44692692318](https://github.com/cflmy/marqdo-cli/actions/runs/44692692318)
and is **red**:
- Linux `fmt + clippy + workspace tests + smoke`: PASS.
- Schema instance validation + Python consumers: PASS.
- macOS `cargo test --workspace`: FAIL —
`plugin_inspect_and_explicit_trusted_invocation` used a Linux `.so` and a
Linux-target manifest.
- Windows `cargo test --workspace`: FAIL — AST golden bytes were CRLF after
checkout while the parser emits LF, producing "0 line(s) differ".
- Linux `fmt + clippy + test`: FAIL —
`strict_mode_enforces_isolation_or_refuses_closed` hit
`bwrap: Creating new namespace failed: Resource temporarily unavailable`.
- macOS / Windows `cargo test --workspace`: FAIL — `E0433` on
`linux::probe()` under `cfg!`.
- Schema instance validation + consumers: PASS.
- Release candidate artifact: SKIPPED (upstream failure; tag/dispatch only).
This checkout has a Gitee remote only and no GitHub Actions credentials. **No
new remote CI URL can be produced from this host.** The local gate below is
real evidence for this machine, not a substitute for the native Windows/macOS
lanes.
This checkout has a Gitee remote and **no GitHub credentials**. No new remote
CI URL can be produced from this host. The evidence below is local/VM evidence
for the convergence commit, not a substitute for the native GitHub lanes.
### Local gate after the stabilization commits
Convergence code/security SHA: **`f075c967f24d885ac42c1e99329bee3498154598`**
(the following docs commit only adds this report and release material; code is
byte-identical).
All commands were run from the repository root with Rust 1.100.0:
### Local static and test gate (Ubuntu 26.04, Rust 1.100.0)
```text
cargo fmt --all -- --check PASS
cargo clippy --workspace --all-targets -- -D warnings PASS
cargo test --workspace PASS
cargo test -p marqdo-cli --test conformance_ast PASS (4 tests)
cargo test -p marqdo-cli --test integration_cli PASS (23 tests)
cargo test -p marqdo-cli --test plugin_worker PASS (4 tests)
cargo test -p marqdo-foreign --test provider PASS (8 tests)
cargo test --workspace PASS (47 binaries,
360 tests, 0 failed)
cargo test --workspace --locked PASS
(47 test binaries, 0 failed; includes 7/7 Linux security tests)
```
Linux G-C security lane (disposable Ubuntu 26.04 VM, ordinary user `dev`):
### G-C security lane (disposable Ubuntu 26.04 VM, ordinary user `dev`)
Source: `git archive f075c96` extracted in the VM (not the dev container).
Host: QEMU/TCG, kernel `7.0.0-43-generic`, `systemd-detect-virt --container`
= `none`, bubblewrap 0.11.1 with `bwrap-userns-restrict`, uid 1000.
```text
crates/marqdo-sandbox/tests/security_linux.rs PASS (7/7)
tests/security/run-lane.sh PASS
./tests/security/run-lane.sh
adversarial helper matrix: 7 passed, 0 failed, 0 ignored
positive CLI contract: PASS
external-execution surfaces refused: PASS
forged active marker rejected: PASS
G-C security lane: PASS
cargo test --workspace --locked (same archive): 47 ok, 0 failed
```
The committed code SHA for this evidence is `b6fc182`; the VM lane was re-run
from a `git archive` of that commit. See
[`evidence/2026-11-20-strict-gc.md`](evidence/2026-11-20-strict-gc.md).
Windows checkout reproduction: a fresh local clone with
`git -c core.autocrlf=true clone ...` kept goldens at LF, preserved the CRLF
byte fixture, and passed all four `conformance_ast` tests. The previous
behavior was reproduced by converting one golden to CRLF: the new diagnostic
reports the first differing byte and "the only difference is line endings".
macOS reproduction is **not possible on this host**; the fix removes the Linux
assumption by generating the manifest from the host triple and building
`solver.dylib` with `-dynamiclib`. It remains UNVERIFIED until the macOS lane
runs.
Logs: [`evidence/2026-11-20-strict-gc-f075c96.md`](evidence/2026-11-20-strict-gc-f075c96.md).
This is still a **local VM lane**, not a self-hosted GitHub Actions security URL.
## 2. Work package status
| # | Work package | State | Evidence |
|---|---|---|---|
| 0 | Windows AST line endings | **Fixed locally, Windows CI UNVERIFIED** | `.gitattributes`, byte diagnostics, CRLF structure test, autocrlf clone PASS |
| 1 | macOS native plugin fixtures | **Fixed locally, macOS CI UNVERIFIED** | host-target manifest + dylib flags, foreign-target/tamper/ABI tests PASS |
| 2 | Status/docs reconcile | **Done** | this file, `COMPATIBILITY.md`, `CHANGELOG.md`, `docs/releasing.md`, `docs/security-gate.md` |
| 3 | Package lock v3 reproducibility | **Implemented; targeted audit pending** | ADR-0016, `cargo test -p marqdo-cli --test package_env` PASS |
| 4 | Analysis/Trace/Tool protocol | **Implemented; consumers PASS locally** | schemas + `tests/consumers/*.py`, `analysis_trace`/`tool_protocol` PASS; `check_never_spawns_a_foreign_kernel` PASS |
| 5 | Foreign pinned environment | **Done (ADR-0020)** | provider tests PASS; env binds interpreter; version drift rejected |
| 6 | ABI session/broker hardening | **Partial** | `fs.read` bridge + budgets + handle tests PASS; Windows ABI unsupported; broader services still `capability.unavailable` |
| 7 | Linux strict probe | **Done** | runtime Bubblewrap probe; unavailable host reports the real reason |
| 8 | Linux strict pure (L1) | **DONE (VM-verified)** | ADR-0021; 7/7 adversarial tests + CLI matrix PASS in disposable Ubuntu VM |
| 9 | Linux strict extensions (L2) | **NOT IMPLEMENTED** | strict rejects Worker/Foreign/native ABI; future ADR needed |
| 10 | Native platform matrix | **Partial** | Linux PASS; macOS/Windows source fixes, remote lanes absent |
| 11 | Independent ecosystem consumers | **Partial** | Analysis/Trace Python consumers PASS; three Tool fixtures not added |
| 12 | RC/docs/artifacts | **Candidate build only** | release docs, checksummed artifact, clean-install smoke; NOT released |
| 0 | Windows AST line endings | Fixed locally; **Windows CI UNVERIFIED** | `.gitattributes`, CRLF fixture |
| 1 | macOS native plugin fixtures | Fixed locally; **macOS CI UNVERIFIED** | host-target manifest + dylib flags |
| 2 | Cross-platform strict `cfg` | **Fixed locally; native CI UNVERIFIED** | `#[cfg]` dispatch + non-Linux regression test |
| 3 | Strict probe→launch race | **Fixed locally; GitHub Linux CI UNVERIFIED** | fail-closed CLI contract + probe-OK/launch-EAGAIN seam test |
| 4 | Parallel security matrix isolation | **Fixed** | unique paths + helper `OnceLock`; 7/7 in parallel workspace run |
| 5 | Markdown/power/control-flow corpus | **Done locally** | 160 AST, 44 semantic, 52 compile-negative, 12 runtime-negative fixtures |
| 6 | Condition header + source spans | **Fixed** | executable tails rejected; literal-inside-backtick mapping tests |
| 7 | Independent Tool consumers | **Done locally** | `sample-view`/`sample-knowledge`/`sample-proof` E2E test |
| 8 | lock v3 reproducibility | **Implemented; tamper test added** | `package_env` PASS |
| 9 | Analysis/Trace/Tool/ABI/Foreign | **Implemented; supported-matrix bounded** | protocol tests, plugin/worker/foreign tests |
| 10 | Linux strict L1 | **VM PASS at f075c96; no CI URL** | 7/7 + positive CLI |
| 11 | Linux strict L2 | **NOT IMPLEMENTED / UNSUPPORTED** | strict rejects Worker/Foreign/native ABI |
| 12 | Windows Native/Worker ABI | **UNSUPPORTED** | no Windows ABI artifact/claim |
| 13 | RC artifact | **Linux candidate only; no release** | clean-install + checksums; macOS not built here |
| 14 | Release approval | **NOT STARTED** | requires owner approval |
## 3. Verified capabilities
- Frozen Markdown profile `marqdo.md.v1` with 150 AST goldens and explicit
- Frozen Markdown profile `marqdo.md.v1` with 160 AST goldens and explicit
CRLF byte semantics.
- Native Rust interpreter with 36 semantic goldens, 48 compile-error fixtures,
- Native Rust interpreter with 44 semantic goldens, 52 compile-error fixtures,
and runtime error source attribution.
- Native `**` power is pinned end-to-end: right associativity, unary-minus
precedence, parentheses, float negative exponents, branch/loop conditions,
Strong+CodeSpan payloads, and inert prose/code fences.
- lock v3 full transitive package-instance graph, verified Git cache, strict
`sync --locked`, atomic `.marqdo/env` materialization; `run`/`check` never
fetch or rewrite the lock.
fetch or rewrite the lock. Locked path content tamper is rejected without
rewriting the lock or environment.
- Analysis v2 (logical identity, resources, byte digest) and Trace v2 (module
ownership, bounds, truncation, stale detection) with independent Python
consumers and JSON Schema validation.
- Tool Protocol v1 with arbitrary command names, SemVer/protocol negotiation,
whole-bundle digest verification, atomic install/remove.
- Tool Protocol v1 with arbitrary command names (no CLI hard-coding),
SemVer/protocol negotiation, whole-bundle digest verification, atomic
install/remove, and three independent Analysis/Trace consumers.
- Plugin ABI v1: native + worker transports, capability broker executing
`fs.read` through `ProjectFs`, per-invocation budgets, host-owned handle
identity, crash/timeout isolation. Not an OS sandbox.
- Foreign kernels: one-shot Python/Node, typed wire, timeout/output caps,
provider and environment digests, and (ADR-0020) environment-bound provider
with interpreter-version verification.
provider/environment digests, environment-bound provider, interpreter
version verification.
- Linux strict L1: Bubblewrap namespace supervisor, runtime probe, fail-closed
launch, child self-verification, 7/7 adversarial matrix.
## 4. Open blockers and limits
1. **No remote CI evidence for this commit.** The repository remote is Gitee;
no GitHub workflow was executed. Windows/macOS fixes are UNVERIFIED there.
2. **G-C is VM-verified but not CI-signed.** Strict L1 passes the full
adversarial matrix in a disposable Ubuntu 26.04 VM (ADR-0021,
[`security-gate.md`](security-gate.md)), and is enabled only when the
runtime Bubblewrap probe succeeds. No remote security-lane URL exists yet,
and strict L2 (Worker/Foreign) is not implemented.
3. Windows Native/Worker ABI is explicitly unsupported (no `cc` toolchain in
the lane); it is not counted as a pass.
1. **No remote CI evidence for the convergence SHA.** Windows/macOS fixes and
the Linux EAGAIN contract are UNVERIFIED on GitHub. G1/G2/G3 remain FAIL
until a real run exists.
2. **G-C is VM-verified but not CI-signed.** No self-hosted security-lane URL
for f075c96 exists.
3. Windows Native/Worker ABI is explicitly UNSUPPORTED (no Windows `cc`
fixture lane, Unix-gated worker tests).
4. `fs.write`, `net.request`, `process`, `clock`, `random` are not executed
broker services; unsupported requests must return `capability.unavailable`.
5. uv/npm/venv provisioning is not implemented. Only a pre-provisioned,
digest-pinned directory with a regular-file executable is supported.
6. Three independent Tool consumers (`sample-view`/`knowledge`/`proof`) are
not yet added; only the portable `benchmark` fixture and Analysis/Trace
Python consumers exist.
7. No release tag/binary publication. A candidate archive and checksums are
produced locally; the release decision is **DO NOT RELEASE** until the
native CI matrix and a security-lane CI URL exist.
6. No Windows/macOS RC archive exists from this host. The release-candidate
workflow can build macOS on GitHub, but it has not run for this SHA.
7. No release tag/binary publication. The release decision is **DO NOT
RELEASE** until the native CI matrix and a security-lane CI URL exist.
## 5. Evidence files
- [`evidence/2026-11-20-local-gate.md`](evidence/2026-11-20-local-gate.md) —
exact local commands, Windows autocrlf reproduction and clean-install smoke.
- [`security-gate.md`](security-gate.md) — the full G-C VM evidence, escape
matrix and strict L1/L2 limits.
earlier local gate and Windows autocrlf reproduction.
- [`evidence/2026-11-20-strict-gc-f075c96.md`](evidence/2026-11-20-strict-gc-f075c96.md)
— current VM lane, environment, full log and workspace test log.
- [`evidence/RELEASE_CANDIDATE_0.1.0_f075c96.md`](evidence/RELEASE_CANDIDATE_0.1.0_f075c96.md)
— G0..G12 matrix and RC decision.
- [`security-gate.md`](security-gate.md) — mechanism, attack matrix and strict
L1/L2 limits.
## 6. Reproduce
@@ -142,7 +146,7 @@ export PATH="$HOME/.cargo/bin:$PATH"
# On hosts with a read-only $HOME, point CARGO_HOME at a writable directory.
cargo fmt --all -- --check
cargo clippy --workspace --all-targets -- -D warnings
cargo test --workspace
cargo test --workspace --locked
cargo build --release --locked -p marqdo-cli
```
@@ -154,3 +158,9 @@ Consumer evidence:
python3 tests/consumers/analysis_consumer.py /tmp/a.json
python3 tests/consumers/trace_consumer.py /tmp/a.json /tmp/t.json
```
G-C (ordinary user, disposable VM only):
```bash
MARQDO_REQUIRE_STRICT_SECURITY=1 ./tests/security/run-lane.sh
```
@@ -0,0 +1,38 @@
Warning: Permanently added '[127.0.0.1]:2222' (ED25519) to the list of known hosts.
== adversarial helper matrix ==
Compiling marqdo-sandbox v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-sandbox)
Finished `test` profile [unoptimized + debuginfo] target(s) in 9.71s
Running tests/security_linux.rs (/home/dev/marqdo-src/target/debug/build/marqdo-sandbox/73fdd38fdea71ea2/out/security_linux-73fdd38fdea71ea2)
running 7 tests
test strict_child_cannot_escape_via_proc_or_privileges ... ok
test strict_child_cannot_reach_the_network ... ok
test strict_child_cannot_read_host_files_but_reads_the_project ... ok
test strict_child_cannot_write_outside_or_into_the_project ... ok
test strict_child_identity_is_unprivileged_and_namespaced ... ok
test strict_supervisor_enforces_fd_and_process_limits ... ok
test strict_supervisor_enforces_memory_cpu_and_output_limits ... ok
test result: ok. 7 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.12s
== build CLI ==
Compiling marqdo-span v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-span)
Compiling marqdo-sandbox v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-sandbox)
Compiling marqdo-package v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-package)
Compiling marqdo-value v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-value)
Compiling marqdo-expression v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-expression)
Compiling marqdo-plugin-abi v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-plugin-abi)
Compiling marqdo-math v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-math)
Compiling marqdo-foreign v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-foreign)
Compiling marqdo-ir v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-ir)
Compiling marqdo-host v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-host)
Compiling marqdo-markdown v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-markdown)
Compiling marqdo-runtime v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-runtime)
Compiling marqdo-frontend v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-frontend)
Compiling marqdo-cli v0.1.0 (/home/dev/marqdo-f075c96/crates/cli)
Finished `dev` profile [unoptimized + debuginfo] target(s) in 1m 49s
strict mechanism: bubblewrap 0.11.1 (user/mount/PID/net/UTS/IPC namespaces + no-new-privs + rlimits)
== positive CLI contract ==
== external-execution surfaces are refused ==
== forged active marker is rejected ==
G-C security lane: PASS
@@ -0,0 +1,114 @@
# G-C strict-isolation evidence — source archive `f075c96` (2026-11-20)
Source SHA: **`f075c967f24d885ac42c1e99329bee3498154598`**
Source archive: `git archive --format=tar.gz --prefix=marqdo-f075c96/ f075c96`
SHA-256 `dc5df444ef6c4f4c073005f355c95bc765c0e4a8c593799e14c22ae10e0520e3`.
Status: **G-C local VM lane PASS**, run as ordinary user `dev` (uid 1000) in a
disposable Ubuntu 26.04 QEMU/TCG VM. **No self-hosted GitHub Actions security
URL exists for this SHA**, so this is not yet a CI-signed G-C.
Full logs:
[`2026-11-20-strict-gc-f075c96.log`](2026-11-20-strict-gc-f075c96.log),
[`2026-11-20-workspace-tests-f075c96.log`](2026-11-20-workspace-tests-f075c96.log).
## Environment
```text
id: uid=1000(dev) gid=1000(dev) groups=1000(dev),27(sudo)
uname: Linux marqdo-security 7.0.0-43-generic #43-Ubuntu SMP ... x86_64
virt: systemd-detect-virt --container -> none (QEMU/TCG guest)
bwrap: bubblewrap 0.11.1 (bwrap-userns-restrict)
cc: cc (Ubuntu 15.2.0-16ubuntu1) 15.2.0
node: v22.22.1
python: Python 3.14.4
rustc: rustc 1.100.0 (0ef1bc386 2026-11-09)
uid_map: 0 0 4294967295
cgroup2: rw,nosuid,nodev,noexec,relatime,nsdelegate,...
```
The lane was extracted from `git archive`, not run from the development
container. `CARGO_TARGET_DIR` pointed at the cached VM target directory; the
lane was fixed to honour that variable rather than hard-coding `target/debug`.
## Lane command
```bash
cd ~/marqdo-f075c96
export PATH=$HOME/.cargo/bin:$PATH
export CARGO_TARGET_DIR=$HOME/marqdo-src/target
MARQDO_REQUIRE_STRICT_SECURITY=1 ./tests/security/run-lane.sh
```
## Result
```text
== adversarial helper matrix ==
running 7 tests
test strict_child_cannot_escape_via_proc_or_privileges ... ok
test strict_child_cannot_reach_the_network ... ok
test strict_child_cannot_read_host_files_but_reads_the_project ... ok
test strict_child_cannot_write_outside_or_into_the_project ... ok
test strict_child_identity_is_unprivileged_and_namespaced ... ok
test strict_supervisor_enforces_fd_and_process_limits ... ok
test strict_supervisor_enforces_memory_cpu_and_output_limits ... ok
test result: ok. 7 passed; 0 failed; 0 ignored
strict mechanism: bubblewrap 0.11.1 (user/mount/PID/net/UTS/IPC namespaces + no-new-privs + rlimits)
== positive CLI contract ==
== external-execution surfaces are refused ==
== forged active marker is rejected ==
G-C security lane: PASS
```
The same archive ran `cargo test --workspace --locked`: **47 test binaries,
0 failed**, including the seven security cases in parallel (the parallel
scratch-path collision is fixed in this SHA).
## Case-level observations
| Attack/surface | Observed inside sandbox |
|---|---|
| identity | `uid=1000`, `pid=1`, `NoNewPrivs: 1`, `CapEff: 0` |
| `/etc/shadow`, `/root`, `$HOME` | `DENIED` |
| `/project/allowed.txt` | `OK` |
| write outside project / into `/project` | `DENIED`; no host file created |
| write `/tmp` | allowed in private tmpfs, host unaffected |
| `/proc/1/root/etc/shadow` | `DENIED` |
| inherited fds above stdio | none |
| `setuid(0)`, `mount`, nested `unshare` | `DENIED` |
| IPv4 connect / raw socket | `DENIED` |
| fd loop | `EMFILE` <= 64 |
| fork loop | bounded below 200 |
| 768 MiB allocation | denied/killed by 512 MiB `RLIMIT_AS` |
| infinite CPU loop | stopped by wall-clock/CPU budget |
| 4 MiB stdout | stopped at the 64 KiB output limit |
## CLI behavior in the VM
```text
run examples/hello.mq.md --mode strict -> 你好,Marqdo (exit 0)
call examples/factorial.mq.md 数学.阶乘 --arg n=5 --mode strict -> 120
run examples/multi-module/index.mq.md --mode strict -> exit 0
run --mode strict --json -> success: true
run --mode strict --allow-foreign -> exit 2
run --mode strict --allow-read/--allow-write -> exit 2
run --mode strict --trust-native-plugin --plugin ... -> exit 2
run --mode strict --kernel ... -> exit 2
run --mode strict --trace-file ... -> exit 2
MARQDO_STRICT_ACTIVE=1 run --mode strict -> exit 1 (not PID 1)
```
## Residual risks and non-claims
* Strict L2 (Worker/Foreign/native ABI inheriting the sandbox) is not
implemented; those flags are rejected. Native ABI is never allowed in strict.
* rlimits are used instead of cgroup-v2 accounting; no cgroup enforcement is
claimed.
* Filesystem TOCTOU is reduced by read-only mounts but not formally proven
against a racing host user.
* Network is denied by the network namespace, not a seccomp syscall allowlist.
* **A remote self-hosted security-lane CI run is still required** before a
release may call G-C CI-verified. `.github/workflows/security-linux.yml`
defines that lane; `MARQDO_REQUIRE_STRICT_SECURITY=1` makes a missing
mechanism fail instead of skip.
@@ -0,0 +1,672 @@
Warning: Permanently added '[127.0.0.1]:2222' (ED25519) to the list of known hosts.
Compiling marqdo-plugin-sdk v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-plugin-sdk)
Compiling marqdo-runtime v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-runtime)
Compiling marqdo-foreign v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-foreign)
Compiling marqdo-cli v0.1.0 (/home/dev/marqdo-f075c96/crates/cli)
Compiling marqdo-package v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-package)
Compiling marqdo-host v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-host)
Compiling marqdo-frontend v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-frontend)
Compiling marqdo-markdown v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-markdown)
Compiling marqdo-ir v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-ir)
Compiling marqdo-sandbox v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-sandbox)
Compiling marqdo-math v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-math)
Compiling marqdo-plugin-abi v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-plugin-abi)
Compiling marqdo-value v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-value)
Compiling marqdo-expression v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-expression)
Compiling marqdo-span v0.1.0 (/home/dev/marqdo-f075c96/crates/marqdo-span)
Finished `test` profile [unoptimized + debuginfo] target(s) in 2m 35s
Running unittests src/main.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/44db2e5fd6960968/out/marqdo-44db2e5fd6960968)
running 2 tests
test args::tests::rejects_non_constant_and_malformed_arguments ... ok
test args::tests::parses_typed_arguments ... ok
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.04s
Running tests/analysis_trace.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/8805c2b5d230266a/out/analysis_trace-8805c2b5d230266a)
running 6 tests
test failed_check_removes_stale_analysis ... ok
test analysis_v2_projects_addressable_formula_and_code_resources ... ok
test analysis_v2_identity_is_location_independent ... ok
test trace_v2_budget_truncates_long_loops_without_failing_the_run ... ok
test trace_records_calls_effects_and_status_without_secrets ... ok
test analysis_is_deterministic_and_has_stable_semantic_nodes ... ok
test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.69s
Running tests/conformance_ast.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/4a86832a684510a9/out/conformance_ast-4a86832a684510a9)
running 4 tests
test byte_diff_helpers_distinguish_line_endings_from_semantics ... ok
test ast_goldens_use_canonical_lf_bytes ... ok
test crlf_fixture_is_preserved_as_bytes ... ok
test ast_goldens_match_the_frozen_profile ... ok
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.56s
Running tests/conformance_semantics.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/775a9ac61fe4e563/out/conformance_semantics-775a9ac61fe4e563)
running 1 test
test semantic_goldens_match_end_to_end_execution ... ok
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 4.21s
Running tests/examples_r6.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/c12f7ae83e17dbbf/out/examples_r6-c12f7ae83e17dbbf)
running 4 tests
test public_schemas_are_valid_json_and_versioned ... ok
test unknown_tool_name_installs_and_dispatches ... ok
test multi_version_packages_run_without_an_implicit_override ... ok
test external_kernel_provider_runs_without_core_changes ... ok
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.98s
Running tests/integration_cli.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/1987735ce297336b/out/integration_cli-1987735ce297336b)
running 20 tests
test condition_parse_diagnostic_points_inside_the_code_span ... ok
test check_exit_codes_and_diagnostics ... ok
test fmt_normalises_line_endings_and_trailing_newlines ... ok
test check_never_spawns_a_foreign_kernel ... ok
test inspect_outputs_ast_units_signatures_ir_and_refs ... ok
test init_creates_a_runnable_project ... ok
test call_executes_examples_with_typed_arguments ... ok
test lock_and_sync_locked_detect_dependency_tampering ... ok
test platform_reports_strict_capability_truthfully ... ok
test multi_module_imports_and_cycles ... ok
test plugin_foreign_target_is_rejected_without_loading ... ok
test examples_run_end_to_end_including_foreign_when_available ... ok
test power_branch_loop_and_inert_code_end_to_end ... ok
test run_json_embeds_program_stdout ... ok
test run_refuses_documents_with_compile_errors ... ok
test runtime_error_in_imported_module_reports_that_module ... ok
test plugin_artifact_tampering_is_rejected_before_load ... ok
test strict_mode_enforces_isolation_or_refuses_closed ... ok
test plugin_abi_major_mismatch_is_rejected_before_execution ... ok
test plugin_inspect_and_explicit_trusted_invocation ... ok
test result: ok. 20 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 5.65s
Running tests/negative.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/c746e5e201607331/out/negative-c746e5e201607331)
running 3 tests
test warning_goldens ... ok
test compile_time_negative_goldens ... ok
test runtime_negative_goldens ... ok
test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.23s
Running tests/package_env.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/ffe0eb54eae24ed0/out/package_env-ffe0eb54eae24ed0)
running 8 tests
test check_does_not_fetch_a_missing_git_revision ... ok
test deleted_lock_edge_fails_check_and_run ... ok
test run_and_check_require_a_lock_when_dependencies_exist ... ok
test lock_records_transitive_closure_and_sync_materializes_it ... ok
test sync_is_strict_and_never_rewrites_the_lock ... ok
test tampered_path_dependency_content_is_rejected_by_locked_sync ... ok
test transitive_path_dependency_resolves_from_its_own_manifest ... ok
test pinned_git_dependency_locks_syncs_and_runs ... ok
test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.04s
Running tests/plugin_worker.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/d45a194c30207bdf/out/plugin_worker-d45a194c30207bdf)
running 4 tests
test worker_capability_request_budget_stops_a_runaway_plugin ... ok
test worker_transport_runs_without_in_process_trust ... ok
test worker_capability_bridge_reads_granted_file_and_denies_without_grant ... ok
test worker_crash_is_isolated_from_the_host ... ok
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.10s
Running tests/properties.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/90e4a82e1e9db60d/out/properties-90e4a82e1e9db60d)
running 2 tests
test narration_removal_preserves_execution ... ok
test pure_prose_edits_do_not_change_executable_ir ... ok
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.64s
Running tests/tool_protocol.rs (/home/dev/marqdo-src/target/debug/build/marqdo-cli/53f82706fffb0ee3/out/tool_protocol-53f82706fffb0ee3)
running 7 tests
test a_root_record_json_in_the_source_package_is_rejected ... ok
test invalid_manifests_and_path_escapes_are_rejected ... ok
test tampered_binary_and_path_hijack_are_not_executed ... ok
test install_dispatch_list_remove_without_core_knowledge_of_the_name ... ok
test protocol_negotiation_accepts_declared_versions_and_rejects_disjoint_ones ... ok
test tampering_any_bundle_file_or_adding_an_unlisted_file_is_rejected ... ok
test independent_sample_tools_consume_public_analysis_and_trace ... ok
test result: ok. 7 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 26.54s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-expression/0c43a9e46f8c64a9/out/marqdo_expression-0c43a9e46f8c64a9)
running 50 tests
test ast::tests::spanless_variants_return_the_empty_span ... ok
test lexer::tests::bad_text_lexing_reports_errors ... ok
test ast::tests::span_accessor_covers_structural_nodes ... ok
test ast::tests::span_serde_round_trips_inside_ast ... ok
test lexer::tests::comments_are_not_supported ... ok
test lexer::tests::text_escapes_decode ... ok
test lexer::tests::number_lexing_rejects_invalid_forms ... ok
test lexer::tests::tokenizes_keywords_separately_from_identifiers ... ok
test lexer::tests::token_remembers_multibyte_spans ... ok
test lexer::tests::tokenizes_operators_and_punctuation ... ok
test lexer::tests::unicode_whitespace_is_skipped ... ok
test parser::tests::assignment_statements_parse ... ok
test parser::tests::additive_and_multiplicative_precedence ... ok
test parser::tests::boolean_and_none_literals_parse ... ok
test parser::tests::calls_parse_positional_and_named_arguments ... ok
test parser::tests::comments_are_not_supported ... ok
test parser::tests::comparisons_are_non_associative ... ok
test parser::tests::call_argument_errors_are_rejected ... ok
test parser::tests::comparisons_bind_looser_than_arithmetic ... ok
test parser::tests::error_spans_point_at_the_offending_token ... ok
test parser::tests::expression_statements_parse ... ok
test parser::tests::expressions_reject_assignment_syntax ... ok
test parser::tests::float_literals_cover_the_grammar ... ok
test parser::tests::integer_literals_are_arbitrary_precision ... ok
test parser::tests::invalid_float_literals_are_rejected ... ok
test parser::tests::invalid_text_literals_are_rejected ... ok
test parser::tests::invalid_statements_are_rejected ... ok
test parser::tests::list_literal_errors_are_rejected ... ok
test ast::tests::ast_types_serde_round_trip ... ok
test parser::tests::list_literals_parse ... ok
test parser::tests::member_and_index_access_chain ... ok
test parser::tests::logical_operators_chain_left_associatively ... ok
test parser::tests::member_and_index_errors_are_rejected ... ok
test parser::tests::parse_arguments_matches_call_argument_semantics ... ok
test parser::tests::parse_arguments_parses_a_leading_argument_list ... ok
test parser::tests::parse_arguments_rejects_invalid_lists ... ok
test parser::tests::power_is_right_associative_and_binds_tighter_than_unary_minus ... ok
test parser::tests::record_literal_errors_are_rejected ... ok
test parser::tests::reserved_keywords_cannot_be_identifiers ... ok
test parser::tests::spans_are_byte_offsets_into_the_payload ... ok
test parser::tests::record_literals_parse ... ok
test parser::tests::text_literals_decode_escapes ... ok
test parser::tests::unary_operators_stack ... ok
test tests::ast_types_are_reachable_from_the_crate_root ... ok
test parser::tests::unicode_identifiers_are_accepted ... ok
test tests::empty_payloads_are_errors ... ok
test tests::parse_error_displays_code_and_span_and_implements_error ... ok
test tests::public_api_parses_expression_and_statement ... ok
test parser::tests::adversarial_inputs_never_panic ... ok
test parser::tests::nesting_depth_limit_is_enforced ... ok
test result: ok. 50 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.20s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-foreign/5ae4cf14f713cad7/out/marqdo_foreign-5ae4cf14f713cad7)
running 2 tests
test tests::languages_are_normalized ... ok
test tests::request_envelope_is_versioned ... ok
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s
Running tests/kernels.rs (/home/dev/marqdo-src/target/debug/build/marqdo-foreign/f8c333b5425a20ed/out/kernels-f8c333b5425a20ed)
running 22 tests
test call_rejects_non_json_stdout ... ok
test call_rejects_stdout_noise ... ok
test call_decodes_a_python_wire_value ... ok
test call_reports_timeout_as_an_error ... ok
test call_rejects_unknown_wire_types ... ok
test error_messages_are_human_readable ... ok
test child_environment_only_keeps_path ... ok
test detect_finds_python_and_node_versions ... ok
test kernel_that_ignores_stdin_does_not_hang ... ok
test non_zero_exit_is_failed_and_keeps_stderr ... ok
test non_zero_exit_is_preserved_by_call ... ok
test oversized_output_is_an_error ... ok
test node_commonjs_syntax_is_supported ... ok
test oversized_stderr_is_an_error ... ok
test relative_writes_happen_in_a_fresh_cleaned_directory ... ok
test node_run_success_captures_stdout ... ok
test call_decodes_a_node_wire_value ... ok
test run_success_captures_stdout ... ok
test unknown_language_is_unavailable ... ok
test timeout_kills_the_kernel ... ok
test stdin_envelope_carries_the_input ... ok
test stdin_envelope_sends_null_for_missing_input ... ok
test result: ok. 22 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.11s
Running tests/provider.rs (/home/dev/marqdo-src/target/debug/build/marqdo-foreign/789d5b736522dbbd/out/provider-789d5b736522dbbd)
running 8 tests
test builtin_provider_cannot_be_bound_to_an_environment ... ok
test duplicate_language_labels_and_wire_versions_are_rejected ... ok
test external_provider_timeout_is_enforced ... ok
test pinned_environment_binds_the_provider_and_is_verified_before_execution ... ok
test provider_outside_pinned_environment_is_rejected ... ok
test external_provider_is_registered_without_core_changes ... ok
test declared_interpreter_version_must_match_the_resolved_program ... ok
test provider_digest_is_verified_at_registration_and_execution ... ok
test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.23s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-frontend/9ee456f4c9890a8d/out/marqdo_frontend-9ee456f4c9890a8d)
running 35 tests
test analyze::tests::invalid_list_header_is_a_hard_error ... ok
test analyze::tests::invalid_strong_payload_is_a_hard_error ... ok
test analyze::tests::maps_countdown_loop ... ok
test calls::tests::builtin_arity_is_checked ... ok
test analyze::tests::maps_recursive_factorial ... ok
test calls::tests::inferred_signature_requires_named_arguments ... ok
test calls::tests::explicit_signature_allows_positional_and_checks_arity ... ok
test calls::tests::unknown_function_is_an_error ... ok
test calls::tests::declared_plugin_export_is_checked_statically ... ok
test infer::tests::explicit_signature_reports_unknown_reads ... ok
test infer::tests::infers_free_reads ... ok
test infer::tests::loop_counter_is_parameter ... ok
test infer::tests::unconditional_entry_assignment_is_local ... ok
test analyze::tests::prose_changes_do_not_change_ir ... ok
test infer::tests::partial_branch_assignment_is_ambiguous ... ok
test payload::extra_span_tests::double_backtick_code_span_maps_inside_both_delimiters ... ok
test payload::extra_span_tests::condition_description_rejects_executable_roles ... ok
test payload::extra_span_tests::condition_literal_span_excludes_backticks_for_one_and_two_ticks ... ok
test payload::extra_span_tests::condition_description_allows_inert_inline_content ... ok
test payload::extra_span_tests::utf8_prefix_and_multiline_payload_map_to_original_bytes ... ok
test payload::tests::payload_and_spans_are_preserved ... ok
test payload::tests::span_maps_across_text_and_code_segments ... ok
test payload::tests::nested_markup_is_a_hard_error ... ok
test signature::tests::parses_untyped_and_empty ... ok
test signature::tests::parses_typed_signature_with_default ... ok
test signature::tests::split_respects_nesting_and_strings ... ok
test signature::tests::rejects_duplicates_and_bad_defaults ... ok
test resources::tests::content_hash_is_stable ... ok
test resources::tests::formulas_and_code_are_attributed_to_functions ... ok
test units::tests::duplicate_function_names_are_errors ... ok
test units::tests::empty_heading_is_an_error ... ok
test units::tests::builds_function_ranges ... ok
test units::tests::executable_outside_function_warns_but_does_not_execute ... ok
test units::tests::root_level_function_without_namespace ... ok
test units::tests::same_name_in_different_namespaces_is_allowed ... ok
test result: ok. 35 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.28s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-host/8baaeabbbfb903b3/out/marqdo_host-8baaeabbbfb903b3)
running 10 tests
test budget::tests::output_and_stack_limits ... ok
test budget::tests::fuel_is_bounded ... ok
test capability::tests::defaults_deny_external_io ... ok
test capability::tests::explicit_grants_are_glob_scoped ... ok
test fs::tests::rejects_traversal_and_absolute_paths ... ok
test fs::tests::reads_and_writes_inside_root ... ok
test capability::tests::broker_enforces_core_grants_and_vendor_policy ... ok
test capability::execute_tests::broker_executes_fs_read_and_reports_unavailable_for_other_capabilities ... ok
test handles::tests::forged_cross_owner_and_expired_handles_are_rejected ... ok
test handles::tests::handles_inside_untrusted_values_are_detected ... ok
test result: ok. 10 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.06s
Running tests/capability_broker.rs (/home/dev/marqdo-src/target/debug/build/marqdo-host/2c4ec063e83a0665/out/capability_broker-2c4ec063e83a0665)
running 1 test
test granted_fs_read_returns_real_bytes_and_denial_is_audited ... ok
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.06s
Running tests/plugin_exports.rs (/home/dev/marqdo-src/target/debug/build/marqdo-host/23f86a8716dc2ffd/out/plugin_exports-23f86a8716dc2ffd)
running 1 test
test every_export_is_registered_by_complete_name_and_conflicts_are_atomic ... ok
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.76s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-ir/0a42156cdf4add17/out/marqdo_ir-0a42156cdf4add17)
running 7 tests
test builtins::tests::lookup_aliases ... ok
test builtins::tests::arity_validation ... ok
test tests::ir_serializes_for_inspect ... ok
test callable::tests::builtins_are_registered_with_aliases ... ok
test callable::tests::external_names_must_be_unique_and_not_shadow_builtins ... ok
test tests::module_lookup_helpers ... ok
test tests::stmt_keeps_spans ... ok
test result: ok. 7 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.05s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-markdown/bdef3a0717e05139/out/marqdo_markdown-bdef3a0717e05139)
running 16 tests
test frontmatter::tests::empty_frontmatter_is_null ... ok
test frontmatter::tests::invalid_yaml_is_reported_not_panicked ... ok
test frontmatter::tests::import_shapes_are_validated ... ok
test frontmatter::tests::parses_import_mapping ... ok
test profile::tests::bom_is_tolerated_and_spans_shift ... ok
test profile::tests::profile_options_are_frozen ... ok
test profile::tests::setext_headings_are_heading_nodes ... ok
test profile::tests::math_nodes_are_native_and_inert ... ok
test profile::tests::parse_maps_core_constructs ... ok
test profile::tests::spans_are_byte_offsets_into_original_source ... ok
test profile::tests::crlf_and_lf_sources_keep_structure_and_byte_spans ... ok
test text::tests::payload_accepts_text_and_code ... ok
test text::tests::payload_rejects_nested_markup ... ok
test text::tests::plain_text_uses_alt_and_label ... ok
test text::tests::normalize_collapses_whitespace_and_nfc ... ok
test profile::tests::tables_and_task_lists_are_supported ... ok
test result: ok. 16 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.15s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-math/1b48f3246f930605/out/marqdo_math-1b48f3246f930605)
running 21 tests
test error::tests::constructors_use_stable_kinds ... ok
test error::tests::display_includes_kind_and_position ... ok
test ast::tests::decimal_and_fraction_rendering_is_deterministic ... ok
test ast::tests::variables_are_collected_without_constants ... ok
test ast::tests::rendering_round_trips_through_parse ... ok
test parse::tests::frac_and_sqrt_become_dedicated_nodes ... ok
test parse::tests::empty_and_malformed_input_is_a_parse_error ... ok
test parse::tests::implicit_multiplication_joins_adjacent_factors ... ok
test parse::tests::left_and_right_parentheses_are_accepted ... ok
test parse::tests::more_than_one_equals_is_a_parse_error ... ok
test parse::tests::excessive_node_count_is_rejected ... ok
test parse::tests::power_is_right_associative ... ok
test parse::tests::precedence_matches_standard_arithmetic ... ok
test parse::tests::single_equals_is_an_equation ... ok
test parse::tests::unary_minus_binds_looser_than_power ... ok
test parse::tests::unknown_characters_are_structured ... ok
test parse::tests::unsupported_commands_are_structured ... ok
test parse::tests::deep_nesting_is_rejected_without_stack_overflow ... ok
test tests::public_api_round_trip ... ok
test parse::tests::numeric_literal_bounds_are_enforced ... ok
test parse::tests::very_long_chain_is_rejected_without_stack_overflow ... ok
test result: ok. 21 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.26s
Running tests/kernel.rs (/home/dev/marqdo-src/target/debug/build/marqdo-math/015a83300848687b/out/kernel-015a83300848687b)
running 37 tests
test division_by_zero_is_a_structured_error ... ok
test constants_and_their_ambiguity_rules ... ok
test division_stays_exact_when_it_reduces_to_an_integer ... ok
test decimal_literals_are_exact_rationals ... ok
test dot_and_times_are_multiplication_operators ... ok
test error_display_includes_kind_and_position ... ok
test exponent_must_be_an_integer ... ok
test equation_solving_requires_exactly_one_affine_occurrence ... ok
test f_equals_m_a_solved_for_a ... ok
test f_equals_m_a_solved_for_f ... ok
test f_equals_m_a_solved_for_m ... ok
test exact_big_integer_arithmetic ... ok
test float_bindings_are_exact_numeric_values ... ok
test frac_is_exact_division ... ok
test huge_exponents_are_bounded_not_hung ... ok
test hostile_inputs_never_panic ... ok
test large_exact_results_are_returned_as_integers ... ok
test left_right_parentheses_and_nesting ... ok
test negative_integer_exponents_are_exact_reciprocals ... ok
test malformed_latex_is_a_parse_error_with_position ... ok
test linear_equation_solving_both_directions ... ok
test non_numeric_bindings_are_type_errors ... ok
test perfect_and_non_perfect_square_roots ... ok
test pi_r_squared_area ... ok
test scientific_notation_and_adjacent_numbers_are_rejected ... ok
test precedence_and_parentheses ... ok
test solve_target_on_the_left_of_a_product ... ok
test substitute_rejects_unsupported_binding_types ... ok
test substitute_leaves_unbound_variables_symbolic ... ok
test substitute_replaces_bindings_deterministically ... ok
test target_errors_are_structured ... ok
test substitution_with_float_and_negative_bindings_reparses ... ok
test undefined_variables_are_reported ... ok
test unicode_identifiers ... ok
test unsupported_constructs_are_structured ... ok
test long_operator_chains_within_the_node_budget_still_evaluate ... ok
test oversized_formulas_are_rejected_with_a_structured_error ... ok
test result: ok. 37 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.52s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-package/3349516ee391f20b/out/marqdo_package-3349516ee391f20b)
running 19 tests
test fileset::tests::excludes_git_and_marqdo_directories ... ok
test fileset::tests::materialization_rejects_a_wrong_expected_digest ... ok
test fileset::tests::digest_encoding_is_unambiguous ... ok
test fileset::tests::materialization_uses_the_same_selection_and_digest ... ok
test fileset::tests::rejects_symlinks ... ok
test fileset::tests::target_directory_is_package_content ... ok
test tests::discovers_project_root ... ok
test tests::git_dependency_requires_rev ... ok
test fileset::tests::rejects_special_files ... ok
test tests::package_resolution_is_stable_across_path_moves ... ok
test tests::lockfile_serialization_is_stable_and_sorted ... ok
test tests::lock_graph_records_transitive_path_closure ... ok
test tests::missing_locked_edge_fails_closed ... ok
test tests::rejects_absolute_dependency_paths ... ok
test tests::root_manifest_edit_invalidates_the_lock ... ok
test tests::parses_manifest_and_validates_sources ... ok
test tests::v1_and_v2_locks_are_rejected_with_a_relock_hint ... ok
test tests::sync_materializes_a_content_checked_environment ... ok
test tests::pinned_git_dependency_resolves_and_reuses_cache ... ok
test result: ok. 19 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.55s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-plugin-abi/90fd8b9216ff8955/out/marqdo_plugin_abi-90fd8b9216ff8955)
running 27 tests
test abi::tests::layout_matches_c_header_on_64_bit ... ok
test abi::tests::negotiation_accepts_same_major_and_clamps_minor ... ok
test abi::tests::negotiation_rejects_other_majors ... ok
test artifact::tests::hex_validation_is_strict ... ok
test abi::tests::metadata_rejects_empty_and_invalid ... ok
test abi::tests::metadata_wire_roundtrip ... ok
test artifact::tests::host_triple_is_parseable_and_matches_word_size ... ok
test artifact::tests::target_triple_parsing_and_word_size ... ok
test error::tests::optional_fields_are_omitted ... ok
test manifest::tests::abi_requirement_caret_semantics ... ok
test artifact::tests::sha256_matches_known_vector ... ok
test error::tests::error_is_serializable_roundtrip ... ok
test manifest::tests::malformed_yaml_is_rejected ... ok
test manifest::tests::duplicate_and_reserved_exports_are_rejected ... ok
test manifest::tests::valid_manifest_parses_and_validates ... ok
test manifest::tests::artifact_resolution_requires_exact_host_target ... ok
test manifest::tests::invalid_fields_are_rejected ... ok
test wire::tests::request_rejects_wrong_version_or_kind ... ok
test manifest::tests::vendor_capabilities_are_validated_but_not_granted ... ok
test wire::tests::response_rejects_ambiguous_payloads_and_accepts_unit ... ok
test wire::tests::response_success_and_error_roundtrip ... ok
test worker::tests::frame_roundtrip_and_clean_eof ... ok
test worker::tests::oversized_and_zero_length_frames_are_rejected ... ok
test worker::tests::partial_header_is_an_error_not_eof ... ok
test worker::tests::truncated_payload_is_rejected ... ok
test wire::tests::request_roundtrip_preserves_type_tags ... ok
test wire::tests::oversized_wire_is_rejected_before_json_parsing ... ok
test result: ok. 27 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.20s
Running tests/loader_unix.rs (/home/dev/marqdo-src/target/debug/build/marqdo-plugin-abi/795cd8c1c3c5ea31/out/loader_unix-795cd8c1c3c5ea31)
running 8 tests
test malformed_manifest_is_rejected ... ok
test example_manifest_parses_and_validates ... ok
test artifact_target_must_match_host ... ok
test missing_artifact_file_fails_resolution ... ok
test missing_entry_symbol_fails_load ... ok
test sha256_mismatch_fails_verification ... ok
test wrong_abi_major_fails_load ... ok
test c_plugin_roundtrips_wire_request ... ok
test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.80s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-plugin-sdk/a8c1ea2d42f1442c/out/marqdo_plugin_sdk-a8c1ea2d42f1442c)
running 4 tests
test tests::initialization_rejects_bad_host_major ... ok
test tests::handler_error_is_written_as_error_envelope ... ok
test tests::registered_export_roundtrips_wire_request ... ok
test tests::macro_exports_and_registers ... ok
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.07s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-runtime/7e824000026d5c21/out/marqdo_runtime-7e824000026d5c21)
running 6 tests
test value_ops::tests::division_by_zero_is_an_error ... ok
test value_ops::tests::floor_division_and_modulo_signs ... ok
test value_ops::tests::division_always_yields_float ... ok
test value_ops::tests::integer_arithmetic_is_exact ... ok
test value_ops::tests::mixed_comparisons_are_exact ... ok
test value_ops::tests::text_concatenation_and_power ... ok
test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s
Running tests/interpreter.rs (/home/dev/marqdo-src/target/debug/build/marqdo-runtime/c19d767cfe9ce07e/out/interpreter-c19d767cfe9ce07e)
running 10 tests
test explicit_signature_supports_positional_arguments ... ok
test fuel_budget_is_enforced ... ok
test builtins_for_conversions_and_length ... ok
test ambiguous_ref_resource_is_an_error ... ok
test non_bool_condition_is_a_runtime_error ... ok
test function_reference_can_be_called ... ok
test loop_prints_and_returns_unit_for_fallthrough ... ok
test recursion_branch_and_return ... ok
test recursion_depth_is_bounded ... ok
test return_inside_loop_propagates ... ok
test result: ok. 10 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.69s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-sandbox/bb72be089115133e/out/marqdo_sandbox-bb72be089115133e)
running 4 tests
test tests::mode_names_round_trip ... ok
test linux::tests::launch_failure_never_falls_back_to_unrestricted_execution ... ok
test tests::strict_refuses_when_isolation_is_unavailable ... ok
test linux::tests::probe_success_then_launch_eagain_is_fail_closed ... ok
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s
Running tests/security_linux.rs (/home/dev/marqdo-src/target/debug/build/marqdo-sandbox/0bce0bec1aee4fec/out/security_linux-0bce0bec1aee4fec)
running 7 tests
test strict_child_cannot_read_host_files_but_reads_the_project ... ok
test strict_child_cannot_write_outside_or_into_the_project ... ok
test strict_child_cannot_reach_the_network ... ok
test strict_child_cannot_escape_via_proc_or_privileges ... ok
test strict_child_identity_is_unprivileged_and_namespaced ... ok
test strict_supervisor_enforces_fd_and_process_limits ... ok
test strict_supervisor_enforces_memory_cpu_and_output_limits ... ok
test result: ok. 7 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 5.18s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-span/320ae70514012226/out/marqdo_span-320ae70514012226)
running 2 tests
test tests::span_merge_and_shift ... ok
test tests::line_index_maps_multibyte_text ... ok
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s
Running unittests src/lib.rs (/home/dev/marqdo-src/target/debug/build/marqdo-value/997168d51ecaf867/out/marqdo_value-997168d51ecaf867)
running 10 tests
test value::tests::big_integers_are_arbitrary_precision ... ok
test error::tests::frames_are_rendered_as_related_diagnostics ... ok
test value::tests::unit_and_none_are_distinct ... ok
test diagnostic::tests::diagnostic_json_shape_is_stable ... ok
test value::tests::display_is_stable ... ok
test value::tests::float_to_bigint_requires_integral_value ... ok
test wire::tests::big_integers_keep_all_digits ... ok
test wire::tests::special_floats_are_tagged ... ok
test wire::tests::scalar_roundtrip ... ok
test wire::tests::untrusted_wire_is_validated ... ok
test result: ok. 10 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.19s
Doc-tests marqdo_expression
running 2 tests
test crates/marqdo-expression/src/lib.rs - (line 64) ... ok
test crates/marqdo-expression/src/lib.rs - (line 11) ... ok
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.08s
Doc-tests marqdo_foreign
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_frontend
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_host
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_ir
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_markdown
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_math
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_package
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_plugin_abi
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_plugin_sdk
running 3 tests
test crates/marqdo-plugin-sdk/src/lib.rs - mq_plugin_entry (line 68) ... ignored
test crates/marqdo-plugin-sdk/src/wire.rs - wire (line 7) ... ignored
test crates/marqdo-plugin-sdk/src/lib.rs - (line 8) ... ok
test result: ok. 1 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 1.28s
Doc-tests marqdo_runtime
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_sandbox
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_span
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
Doc-tests marqdo_value
running 0 tests
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
@@ -0,0 +1,141 @@
# Marqdo CLI release candidate — 0.1.0 / `f075c96`
```text
Source SHA: f075c967f24d885ac42c1e99329bee3498154598
Workspace version: 0.1.0
Intended release target: Linux x86_64 (local RC); macOS/Windows UNVERIFIED
GitHub master: c9f01aa8fb127a2dcc125103ef794832984774ad (unchanged)
Latest public CI: https://github.com/cflmy/marqdo-cli/actions/runs/44692692318 (FAILED)
```
This report records the convergence work and the honest release gate. The
checkout used for this run has a Gitee remote and no GitHub credentials, so no
new remote CI URL can be produced here. The code is ready for a CI run; the
release decision is **DO NOT RELEASE** until the hard gates below have remote
PASS evidence.
## 1. What changed at this SHA
- macOS/Windows `E0433`: strict detection now dispatches through `#[cfg]`
functions, so `linux::probe()` is not type-checked on non-Linux targets.
A non-Linux regression test asserts `strict=unavailable` and fail-closed
refusal.
- Linux strict probe→launch race: `inspect --platform` is a snapshot, not a
promise. The CLI test now accepts only success with correct output or a
non-zero, diagnosed, fail-closed refusal with no project stdout. A sandbox
seam test injects `probe OK → spawn EAGAIN` and proves `LaunchFailed`,
never unrestricted execution.
- Parallel security matrix: unique per-test project/outside paths and a
once-compiled helper, so `cargo test --workspace` can run the 7 adversarial
cases in parallel.
- Condition headers: executable `Strong`/`Emph` in the description remainder
are a hard `markdown.invalid_construct` error; condition expression spans are
mapped from the literal bytes inside the backticks.
- Power/control-flow corpus: 10 AST, 8 semantic, 4 compile-negative and 2
runtime-negative fixtures; CLI E2E branch/loop/inert tests.
- Generic Tool protocol: independent `sample-view`, `sample-knowledge` and
`sample-proof` Analysis/Trace consumers, installed and dispatched by name
with no CLI hard-coding.
- Locked path-dependency tamper test; G-C lane honours `CARGO_TARGET_DIR`.
## 2. Local static and test evidence
Ubuntu 26.04 development container, Rust 1.100.0:
```text
cargo fmt --all -- --check PASS
cargo clippy --workspace --all-targets -- -D warnings PASS
cargo test --workspace --locked PASS
cargo build --release --locked -p marqdo-cli PASS
```
G-C VM, source archive `f075c96`, ordinary user `dev` (uid 1000):
```text
MARQDO_REQUIRE_STRICT_SECURITY=1 ./tests/security/run-lane.sh
adversarial 7/7 PASS
positive CLI contract PASS
external surfaces refused PASS
forged marker rejected PASS
G-C security lane: PASS
cargo test --workspace --locked (same archive) PASS (47 binaries, 0 failed)
```
See [`2026-11-20-strict-gc-f075c96.md`](2026-11-20-strict-gc-f075c96.md).
## 3. Release candidate artifact
```text
archive: marqdo-0.1.0-x86_64-unknown-linux-gnu.tar.gz
sha256: 4e1020735bd7f5bd590b5472e48ed431b1365641b8ac6574254dd963be8cc285
contents: marqdo binary, MIT/Apache licenses, CHANGELOG, COMPATIBILITY,
README, spec/, examples/, tests/consumers/, selected docs,
SOURCE_SHA, rustc -vV
```
Clean-install smoke from an extracted external directory:
```text
sha256sum -c marqdo-0.1.0-x86_64-unknown-linux-gnu.tar.gz.sha256 -> OK
./marqdo --version -> marqdo 0.1.0
./marqdo --help -> ok
./marqdo run examples/hello.mq.md -> 你好,Marqdo
./marqdo check examples/multi-module/index.mq.md --refs --emit ... -> ok
analysis_consumer.py -> functions=2
trace_consumer.py -> events=7
validate_protocols.py -> validated analysis/trace/tool
```
No Windows or macOS archive was produced from this host. The `release-candidate`
GitHub job can build macOS, but it has not run for this SHA.
## 4. Gate matrix
| Gate | Status at f075c96 | Evidence / next action |
|---|---|---|
| G0 same SHA, version, changelog | **PASS (local)** | source archive `f075c96`; version 0.1.0; changelog updated; remote master still `c9f01aa` |
| G1 Linux CI fmt/clippy/workspace/smoke | **FAIL — no remote same-SHA run** | local PASS; push to GitHub and let `CI` run |
| G2 Windows workspace + portable smoke | **FAIL — no remote native run** | cfg fix is local; needs Windows CI |
| G3 macOS workspace + portable smoke | **FAIL — no remote native run** | cfg fix is local; needs macOS CI |
| G4 public Schema + consumers | **UNVERIFIED at same SHA** | local schema/consumer PASS; remote job last passed at `c9f01aa` |
| G5 Markdown AST/power new semantics | **PASS (local)** | 160 AST + 44 semantic fixtures; CLI E2E |
| G6 negative/source-span/inert tests | **PASS (local)** | 52 compile-negative, 12 runtime-negative, span unit/CLI tests |
| G7 lock v3 + package environments | **PASS (local Linux)** | package_env, Git/path tamper, locked sync; Windows/macOS UNVERIFIED |
| G8 Tool/ABI/Worker/Foreign matrix | **PASS (local Linux, bounded)** | generic Tool + 3 consumers; native/worker/foreign tests; Windows ABI UNSUPPORTED |
| G9 Linux strict L1 G-C | **PASS (local VM); CI UNVERIFIED** | 7/7 + CLI; needs self-hosted security-lane URL |
| G10 RC build/checksum/clean install | **PASS (Linux only)** | artifact + sha256 + external clean-install smoke |
| G11 docs/version/license/compatibility | **PASS** | STATUS, COMPATIBILITY, CHANGELOG, security evidence |
| G12 owner approval + tag/Release | **NOT STARTED** | requires explicit project-owner approval |
## 5. Decision
**DO NOT RELEASE.**
Reason: G1/G2/G3 have no real GitHub PASS at this SHA, G4 is not re-proven at
this SHA, and G9 has no self-hosted security-lane URL. Local VM and container
evidence cannot be substituted for those lanes.
## 6. Next actions (in order)
1. Push `f075c96` or its docs-complete successor to GitHub `master` and wait
for all `CI` jobs (Linux, macOS, Windows, schema/consumers) to pass.
2. Register the disposable security VM as a self-hosted runner with labels
`[self-hosted, linux, marqdo-security]`, then run
`.github/workflows/security-linux.yml` at the same SHA. Record the URL and
7/7 result.
3. Run the `release candidate artifact` job at the same SHA; download and
verify the Linux/macOS archives and checksums in a clean directory.
4. Only after G1–G11 have same-SHA URLs, ask the project owner for explicit
approval, then create the version tag and publish the release.
## 7. Known limitations / rollback
- Linux strict L1 only; strict L2 (Worker/Foreign/native) is refused.
- Windows Native/Worker ABI is unsupported.
- macOS/Windows strict is fail-closed, not implemented.
- Foreign requires a pre-provisioned, digest-pinned environment; no uv/npm/venv
auto-provisioning.
- The RC has no signed provenance/SLSA attestation yet; checksums are the only
integrity artifact.
- Rollback is to keep the previous stable release (none has been published) and
withdraw the candidate tag/artifact; no data migration is involved.
+17 -7
View File
@@ -6,9 +6,9 @@ required CI matrix on the exact commit.
## 1. Preconditions
- All applicable G-B gates in
[`STABLE_RELEASE_IMPLEMENTATION.md`](STABLE_RELEASE_IMPLEMENTATION.md) §16.1
are PASS with a CI URL for the same SHA.
- All applicable gates in
[`NEXT_RELEASE_CI_CONVERGENCE.md`](NEXT_RELEASE_CI_CONVERGENCE.md) §7 are
PASS with a CI URL for the same SHA.
- G-C is either PASS with a security-lane URL, or the release explicitly says
strict isolation is unavailable. Never claim G-C from local root/sudo runs.
- `git status --porcelain` is clean apart from intended release files.
@@ -27,7 +27,8 @@ cargo build --release --locked -p marqdo-cli
## 3. Artifact creation
The artifact is a tarball containing the binary, license, README, changelog,
compatibility matrix and specs. Build it from the repository root:
compatibility matrix, specs, examples and the public protocol consumers. Build
it from the repository root:
```bash
version="$(cargo metadata --no-deps --format-version 1 \
@@ -38,7 +39,9 @@ rm -rf "$stage"
mkdir -p "$stage"
install -m 0755 target/release/marqdo "$stage/"
cp LICENSE* CHANGELOG.md COMPATIBILITY.md README.md "$stage/"
cp -r spec "$stage/"
cp -r spec examples "$stage/"
mkdir -p "$stage/tests"
cp -r tests/consumers "$stage/tests/"
tar -C dist -czf "dist/marqdo-${version}-${target}.tar.gz" "$(basename "$stage")"
sha256sum "dist/marqdo-${version}-${target}.tar.gz" \
| tee "dist/marqdo-${version}-${target}.tar.gz.sha256"
@@ -62,8 +65,10 @@ cd "$tmp"/marqdo-*
./marqdo call examples/factorial.mq.md 阶乘 --arg n=5
./marqdo check examples/multi-module/index.mq.md --refs --emit analysis.json
./marqdo run examples/multi-module/index.mq.md --trace-file trace.json
python3 /path/to/tests/consumers/analysis_consumer.py analysis.json
python3 /path/to/tests/consumers/trace_consumer.py analysis.json trace.json
python3 tests/consumers/analysis_consumer.py analysis.json
python3 tests/consumers/trace_consumer.py analysis.json trace.json
python3 tests/consumers/validate_protocols.py analysis.json trace.json \
examples/tool-unknown-name/tool.json
```
Also verify the SHA-256 file before extraction:
@@ -87,6 +92,11 @@ git push origin "v${version}"
Do not force-push, do not tag a red commit, and do not upload macOS/Windows
binaries unless those platform lanes actually passed.
The current `0.1.0` candidate at `f075c96` is **not** releasable: it has local
Linux/VM evidence and a checksummed Linux RC, but no same-SHA GitHub CI or
self-hosted G-C URL. See
[`evidence/RELEASE_CANDIDATE_0.1.0_f075c96.md`](evidence/RELEASE_CANDIDATE_0.1.0_f075c96.md).
## 6. Rollback
- If an asset fails verification: delete the release asset and keep the tag
+11 -1
View File
@@ -1,7 +1,11 @@
# Linux strict-isolation security gate (G-C) — evidence
Status: **PASSED in a disposable Ubuntu VM; Linux strict is enabled only when
the runtime probe succeeds.** Code commit: `b6fc182`. No remote security-lane CI URL exists yet.
the runtime probe succeeds.** Current convergence evidence is for source
archive `f075c96` (see
[`evidence/2026-11-20-strict-gc-f075c96.md`](evidence/2026-11-20-strict-gc-f075c96.md));
the earlier `b6fc182` run is retained below as history. **No remote
self-hosted security-lane CI URL exists yet.**
Snapshot: workspace `0.1.0`, 2026-11-20. See ADR-0021.
This report separates three things that are often conflated:
@@ -70,6 +74,12 @@ The helper is `tests/security/escape_helper.c`, launched through the **same**
lane is `tests/security/run-lane.sh` and the test is
`crates/marqdo-sandbox/tests/security_linux.rs`.
At `f075c96` the same lane was re-run from a clean `git archive` in the
disposable VM as `dev` uid 1000: 7/7 adversarial tests, positive CLI contract,
external-surface refusals, and forged-marker rejection all PASS. The command
now honours `CARGO_TARGET_DIR`; the parallel workspace run also passes all
seven security tests.
```text
7/7 tests PASS:
strict_child_identity_is_unprivileged_and_namespaced
@@ -8,6 +8,7 @@ product.
Build and install (any OS with a Rust toolchain):
```sh
mkdir -p /tmp/marqdo-tool-sample-knowledge/bin
rustc --edition 2021 examples/tool-samples/sample-knowledge/src/main.rs \
-o /tmp/marqdo-tool-sample-knowledge/bin/sample-knowledge
cp examples/tool-samples/sample-knowledge/tool.json /tmp/marqdo-tool-sample-knowledge/
@@ -8,6 +8,7 @@ product.
Build and install (any OS with a Rust toolchain):
```sh
mkdir -p /tmp/marqdo-tool-sample-proof/bin
rustc --edition 2021 examples/tool-samples/sample-proof/src/main.rs \
-o /tmp/marqdo-tool-sample-proof/bin/sample-proof
cp examples/tool-samples/sample-proof/tool.json /tmp/marqdo-tool-sample-proof/
@@ -8,6 +8,7 @@ product.
Build and install (any OS with a Rust toolchain):
```sh
mkdir -p /tmp/marqdo-tool-sample-view/bin
rustc --edition 2021 examples/tool-samples/sample-view/src/main.rs \
-o /tmp/marqdo-tool-sample-view/bin/sample-view
cp examples/tool-samples/sample-view/tool.json /tmp/marqdo-tool-sample-view/
+10 -1
View File
@@ -113,7 +113,14 @@ Each item must:
2. have a `paragraph` as its **first** block;
3. have an inline `code` node as the **first inline node** of that paragraph;
its literal is the condition expression;
4. any remaining inline content in that first paragraph is prose.
4. any remaining inline content in that first paragraph is **inert prose**
(`text`, inline `code`, links, images, math, HTML, strikethrough and line
breaks). An executable `Strong`/`Emph` there is a hard
`markdown.invalid_construct` error, not silently ignored description.
The condition literal is mapped from the bytes **inside** the code-span
backticks. Delimiters are part of the Code node's source span, but condition
expression spans and parse diagnostics point at the literal.
The item's remaining blocks form the branch body. Items are tested in AST
order; the first condition evaluating to `true` is executed once. No match
@@ -235,6 +242,8 @@ The following are compile errors with source spans, never prose fallbacks:
* nested `strong`/`emph`/link inside an executable payload;
* ordered list item whose first inline node is not a code span condition;
* bullet list item without a condition header;
* executable `Strong`/`Emph` in the description remainder of a condition
paragraph (it would otherwise be swallowed);
* non-`Bool` condition (detected statically when the type is known, else at
runtime as `runtime.type`);
* block quote that is not exactly one complete call;
+17 -4
View File
@@ -14,12 +14,12 @@ no test commits, copies or rewrites files outside a throw-away temp directory.
```text
tests/
├── conformance/
│ ├── ast/ 150 × *.mq.md + sibling *.ast.json goldens
│ └── semantics/ 36 × *.mq.md + sibling *.expected.json goldens
│ ├── ast/ 160 × *.mq.md + sibling *.ast.json goldens
│ └── semantics/ 44 × *.mq.md + sibling *.expected.json goldens
├── negative/
│ ├── *.mq.md 48 × compile-time hard-error fixtures + *.expected.json
│ ├── *.mq.md 52 × compile-time hard-error fixtures + *.expected.json
│ ├── warnings/ 6 × inert-executable warning fixtures + *.expected.json
│ └── runtime/ 10 × runtime-error fixtures + *.expected.json
│ └── runtime/ 12 × runtime-error fixtures + *.expected.json
└── README.md this file
```
@@ -56,6 +56,19 @@ cargo test -p marqdo-cli --test properties
cargo test -p marqdo-cli --test integration_cli
```
The power/control-flow slice is pinned by:
* AST fixtures 151–160: code-span vs tight/spaced `**`, multi-backtick spans,
inert prose/code fences, branch/loop headers and math blocks;
* semantic fixtures 37–44: right-associative power, unary-minus precedence,
parentheses, float negative exponents, Strong+CodeSpan assignment, branch,
loop and inert prose;
* negative fixtures 49–52: missing power/condition operands and executable
nodes swallowed by a condition paragraph;
* runtime negatives 11–12: negative integer exponent and exponent budget;
* `crates/marqdo-frontend/src/payload.rs` span tests and
`crates/cli/tests/integration_cli.rs` CLI span/E2E assertions.
Fixture discovery is directory-based and sorted by file name, so the order is
deterministic on every platform. The AST and semantic suites fail loudly if the
corpus shrinks below 100 / 25 fixtures, and the negative suite fails below 30