feat(cli): route print through the bundled stdlib (Compat-B)

- Add Program.print_override and lazy stdlib injection: only programs that
  call 打印/print load std:io and route to io.写值 (ADR-0030).
- Record resource.open/read/write/close as trace effect events so the stdlib
  print path is auditable; HostOp::Print no longer runs on that path.
- Permit only stdio/stdout (existing console) in strict; file/network
  resource opens remain permission.denied.
- Exclude std: modules from the human-readable checked-module count; update
  the Python analysis consumer to accept std: instance ids.
- Add print-routing and strict-console tests, ADR-0030 and evidence.

Gate: fmt, clippy -D warnings, 418 tests passed.
This commit is contained in:
cflmy committed 2026-11-20 22:10:59 +08:00
1 parent e119727c5d
commit e26971f489
21 files changed
+560 -36

No files matched your search

+8 -5
View File
@@ -103,9 +103,12 @@ resource.close(handle: Resource)
caller constructs a `Capabilities` set with a matching `fs_read` pattern.
Network/process resources do not exist in this slice.
* `project`: stdio stdout and explicitly granted project-file reads.
* `strict`: **every** `resource.open` is refused with `permission.denied`.
Strict L1 is unchanged and no new provider is silently enabled. A future ADR
must add real OS-level evidence before strict accepts streams.
* `strict`: only the existing **stdio/stdout** console channel is permitted
(`resource.open("stdio","stdout",["write"])`, console grant); every file,
network, process or device resource open is refused with
`permission.denied`. This is the existing console capability reached through
the resource ABI, not a new OS capability (ADR-0030). Strict L1 is otherwise
unchanged.
* Manifest capability requests are not grants; only the operator-provided
effective set is consulted.
@@ -143,8 +146,8 @@ file bytes or handles.
process, clock, random or device operations.
* No generic third-party provider registration; CM-A09 remains open.
* No zero-copy, async, deadlines per stream, seek or shared memory.
* No strict-mode resource support and no claim that host-in-process I/O is a
sandbox.
* No strict-mode file/network/process resource support and no claim that
host-in-process I/O is a sandbox; strict stdio is the existing console.
## Compatibility
@@ -33,7 +33,8 @@ resource.open("file", "<project-relative path>", ["write"]) -> Resource
* `project` mode only, with an `fs_write` glob grant matching the normalised
relative path.
* `pure` still refuses file resources by mode check; `strict` still refuses
every `resource.open` (CM-A11 fail-closed).
file resources (only the existing stdio console channel is permitted,
ADR-0030).
* Absolute paths, `..`, symlink escapes and non-regular targets are rejected by
`ProjectFs` exactly like the existing `fs.write`.
* A writer owns an in-memory byte buffer. `resource.write` appends bytes and
+101
View File
@@ -0,0 +1,101 @@
# ADR-0030: Compat-B `打印` routing through the bundled stdlib
Status: **Accepted**
Date: 2026-11-20
Relates to: `docs/MARQDO_COMPUTE_CORE_STDLIB_MIGRATION_RFC_0002.md`
(CM-A07, CM-A11, P3), ADR-0027, ADR-0028, ADR-0029
Specs: `spec/evaluation.md`, `spec/security.md`
Evidence: `docs/evidence/2026-11-20-print-compat-b.md`
## Context
ADR-0027 implemented Compat-A: the stdlib print chain exists behind explicit
imports, but builtin `打印` still runs `Value::display()` in Rust. ADR-0029
removed the formatting gap: the stdlib can reproduce `打印` for all 15 value
types. ADR-0028 made a verified stdlib bundle available offline. The remaining
Compat-B requirement is to route the **same name** to one implementation and
prove the old callers behave, before Compat-C removes the legacy formatting
path.
## Decision
### 1. One program-level override
`Program` gains an optional `print_override: FunctionId`. The loader sets it to
the bundled `io.写值` function when:
* the verified bundle is available, and
* at least one loaded module calls `打印` or `print` as a bare builtin call.
The stdlib `io` module is injected only in that case. A program without print
calls keeps the same module graph and Analysis output as before. If no bundle
is available, the loader leaves the override unset and the legacy builtin is the
fallback; an official release always ships the bundle.
### 2. Resolution order
At a call site, name resolution is:
1. a local variable holding a callable (`f = ...; f(x)`) — local shadow wins;
2. `打印`/`print` with `Program.print_override` set — call the stdlib function;
3. the existing builtin table;
4. a same-module/imported function declaration.
This is a controlled compatibility route for exactly two names. It is not a
general name-shadowing mechanism and does not add a new namespace.
### 3. Runtime behavior
The stdlib function is a normal Marqdo function: its source spans, fuel,
frames and trace are the stdlib file's. It formats through
`stdlib/format.mq.md`, encodes UTF-8 through `stdlib/text.mq.md`, and writes
through `resource.open("stdio","stdout",["write"])` plus `resource.write`.
`打印` still returns `Unit` and appends exactly one newline.
The console/output-budget checks remain in the host: the stdlib path uses the
same console grant and output budget as the legacy path. Resource operations
are recorded as Trace `effect` events (`resource.open/read/write/close`); the
legacy `HostOp::Print` effect does not appear in Compat-B runs.
### 4. Strict mode and CM-A11
Strict L1 already permits the console channel. Compat-B needs stdout writes in
strict so the same print implementation works there. Therefore strict mode
permits **only** `stdio/stdout` write with the console grant; project-file,
network, process and device resource opens remain refused, and no new provider
is enabled. This is the existing console capability reached through the
resource ABI, not a new OS capability. File resources stay fail-closed.
### 5. Compatibility and fallback
* Output for all 15 value types is the ADR-0029 formatter and is
comparison-tested against the legacy path.
* Trace shape changes from one `print` effect to resource effects. Consumers
that hard-code `operation == "print"` must accept `resource.*`; this is a
documented P3 migration.
* `check`/`analysis` include stdlib modules for print-using programs only.
The human-readable `checked N module(s)` count excludes stdlib modules so
bundle internals do not change the project count.
* Compat-C (removing `HostOp::Print` and `Value::display()` formatting from the
public builtin) is **not** part of this ADR. The legacy path remains the
fallback when no bundle exists.
## Non-goals
* No routing for `长度`, `类型`, `文本`, table or any other builtin.
* No bundled stdlib upgrade workflow; the bundle version stays `0.1.0`.
* No removal of `HostOp::Print`, `Value::display()` or the legacy builtin.
* No new OS isolation claim; strict still refuses file/network/process
resources.
* No guarantee for a CLI installed without `stdlib/`; that configuration keeps
legacy print and is not the official layout.
## Verification
`docs/evidence/2026-11-20-print-compat-b.md` records:
* a print program whose trace contains `resource.write` and no `print` effect;
* all semantic goldens and CLI print tests passing byte-identically;
* lazy stdlib injection: no print → no stdlib modules; print → stdlib modules
and `project.stdlib` identity;
* host tests for strict stdio-only permission and file refusal.