Files
marqdo-cli/CHANGELOG.md
T
cflmy 69a28ed1e1 feat(stdlib): add table GFM rendering and an HTTP/1.1 parser fixture
- stdlib/data/table.mq.md adds 连接片段/连接行/转义单元格/转GFM; logical Table
  renders as a GFM pipe table with pipe/newline escaping (CM-A08 P4 gate).
- Add semantic fixture 71: HTTP/1.1 status/header/chunked response parsing and
  request construction in .mq.md; no host HTTP/TLS implementation (C10).
- Update table CLI test, golden count 71, specs/status/README/changelog and
  evidence.

Gate: fmt, clippy -D warnings, 426 tests passed; bundle digest refreshed.
2026-11-20 23:10:19 +08:00

13 KiB
Raw Blame History

Changelog

All notable changes to the Marqdo CLI workspace. The workspace version is currently 0.1.0; no release tag has been cut. Dates use ISO 8601.

Unreleased (stabilization candidate)

Added

  • .gitattributes byte-identity policy: canonical LF for text files and one explicit CRLF byte fixture for Markdown line-ending semantics.

  • Byte-level AST golden diagnostics: first differing byte, byte windows and an explicit "only line endings differ" classification.

  • Cross-platform Native ABI test fixtures generated for the host target, plus no_matching_artifact, artifact-tamper and ABI-major-mismatch negative tests.

  • Host-specific example plugin builds under examples/plugin-*/build/; the tracked manifests are never rewritten by tests or CI.

  • Foreign pinned-environment binding (ADR-0020): an external provider's entrypoint must resolve inside environment_dir, built-in providers cannot be bound to one, the child gets only the environment executable directory on PATH, and interpreter_version is verified before execution.

  • Linux strict isolation (ADR-0021): a Bubblewrap supervisor with user/mount/ PID/net/UTS/IPC namespaces, no-new-privs, capability dropping, a read-only /project, rlimits, output/wall-clock supervision and child self-verification. Probe-gated; L1 supports pure Marqdo + Math only.

  • Adversarial G-C lane (tests/security/escape_helper.c, crates/marqdo-sandbox/tests/security_linux.rs, tests/security/run-lane.sh) covering filesystem, /proc, network, process, privilege and resource escapes. 7/7 PASS in a disposable Ubuntu 26.04 VM.

  • Release documentation (COMPATIBILITY.md, docs/releasing.md) and the 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.

  • Addressable Memory (RFC-CM-0001 P1, ADR-0022): an opaque run-scoped region value with 64 KiB zero-filled pages, grow-only capacity, bounded byte read/write, memory.new/size/grow, and m[offset] = byte. Copies share the region identity; different regions never alias. Default budgets are 64 MiB committed per run and 16 MiB per region; --max-memory and --max-region-memory override the budgets on run/call.

  • Memory-index place assignment in the expression/IR model (AssignTarget::Index / PlaceTarget::Index) with the frozen evaluation order base, index, RHS and no change to List/Map/Table value semantics.

  • Stable runtime.memory error category for invalid/stale handles and page counts; runtime.index for offsets, runtime.type for operand/byte-range errors, resource.exhausted for budgets.

  • Analysis v2 projects an index assignment target as an additive {"kind":"index",...} object while local targets keep the v1 string.

  • examples/memory-buffer.mq.md, semantic goldens 45–47 and runtime negatives 13–18 cover sharing, isolation, growth, zero-fill, offsets, byte range, non-Memory index assignment, declared-max growth and negative page counts.

  • Partial RFC-CM P2 closure evidence (.mq.md only, no Rust algorithm additions): fixtures 48–50 implement a big-endian u32 codec, a bump allocator, node allocation and a linked-list traversal/sum, and prove that a failed allocation leaves the allocator header unchanged. This is not the full P2 gate: a separate stdlib package and a frozen-CLI closure lane are still missing.

  • Memory handles are never transferable: to_wire emits a bare rejected tag, from_wire fails, and every host/plugin/foreign call rejects arguments that contain a handle.

  • Minimal structure observation/construction (RFC-CM CM-A04, ADR-0023): list.append/concat, map.keys/has/get/set/remove, text.codepoints/from_codepoint, and table.columns/row/cell. All are pure, value-returning builtins; inputs are never mutated and no new syntax is added. Semantic fixtures 51–53 and runtime negatives 19–21 cover value semantics, Unicode scalars, table projection, wrong types and bounds.

  • .mq.md table filter (过滤) built on table.columns/row/cell, list.append and 表, with a semantic golden showing the input table is unchanged (fixture 54). No table.filter/HostOp is added.

  • Recoverable Result convention (RFC-CM CM-A05, ADR-0024): an ordinary Map {ok:true,value} / {ok:false,code,message[,details]}, constructed by .mq.md functions, with expected failures recoverable and language invariant violations still fatal. No Result type, exception or try/catch syntax.

  • .mq.md first-fit free-list allocator (fixture 55): 8-byte block headers, free-block reuse, block splitting, address-adjacent coalescing (tested in a second region whose bump space is exhausted, so only a merged block can satisfy the request), non-overlapping reused payloads and alloc.no_space returned as a Result without aborting the run. No allocator Rust builtin or HostOp was added.

  • .mq.md P2 closures beyond the first slice: binary tree insert/in-order traversal (56), CSV row splitter (57), open-addressing hash table (58), UTF-8 encoder/validating decoder (59–60), graph adjacency list (61), recursive JSON parser for arrays/objects/strings (62), stack bytecode VM (63), 16/32/64-bit + LEB128 codecs (64), table sort (65), table equality join (67) and quoted/escaped CSV (68). No Rust algorithm, container or parser was added for these; algorithms remain .mq.md code. JSON float parsing and a compact formatter are included.

  • Resource/stream first slice (RFC-CM CM-A06, ADR-0025): opaque host-issued Resource handles, resource.open/read/write/close, bytes transferred through Memory, stdout write in pure/project under console/output budgets, granted project-file chunked read in project, EOF as success, closed-handle Result errors, forged-handle fatal errors, wire/host-boundary rejection, and strict refusing every resource open. Semantic fixture 66 and host/CLI tests.

  • Buffered project-file write streams (ADR-0026): file + ["write"] requires an fs_write grant, buffers in the host, flushes only on close, never flushes on drop, has an 8 MiB ceiling, and is refused in pure/strict. A .mq.md chunked copy loop reproduces a file without a Rust copy_file.

  • Compat-C print completion (ADR-0031): remove HostOp::Print and StdHost::op_print; the print builtin only routes to bundled io.写值 or returns kernel.unavailable; print traces contain only resource effects.

  • Compat-B 打印/print routing (ADR-0030): a Program.print_override sends print calls to bundled io.写值 when the program uses them; stdlib modules are injected lazily, resource operations appear as trace effects and the legacy HostOp::Print path does not run. Strict permits only stdio/stdout (existing console), not file/network resources. No Compat-C removal yet.

  • Full-value stdlib formatting (ADR-0029): FunctionRef module/name and PluginHandle plugin/export read-only members (host id private); stdlib/format.mq.md now reproduces builtin 打印/repr for all 15 value types, including Formula/Code and UTF-8 byte length for Code.

  • CM-A10 bundled stdlib (ADR-0028): a fixed stdlib/ tree with format/text/io, explicit std:<path> imports, SHA-256 digest verification before module loading, a std:marqdo-stdlib@<version>#<digest> package instance id, project.stdlib in Analysis v2, and test coverage for std: resolution, tamper rejection and frozen-binary hash equality.

  • CM-A08 GFM rendering: stdlib/data/table.mq.md::转GFM renders a logical Table as a GFM pipe table with cell escaping; CLI test added.

  • P5 HTTP parser fixture (C10): semantic fixture 71 parses HTTP/1.1 status, headers and chunked transfer bodies and builds requests in .mq.md; no host HTTP/HTTPS implementation is claimed.

  • P5 blocking TCP endpoint v1 (ADR-0034): resource.open("tcp", ip:port, ["read","write"]) with exact IP-literal grants, project mode only, 5s connect/read/write timeouts; no DNS/TLS/UDP/server sockets. CLI --allow-net <IP>; loopback host and .mq.md echo tests. Strict/pure refuse.

  • CM-A09 in-process ResourceProvider v1 (ADR-0033): ResourceProvider trait, StdHost::register_resource_provider, per-kind registration with protocol version/conflict checks and Capabilities.allow_resource_provider; provider streams reuse resource.* and are project-mode only. A mock read/write provider test runs without parser/runtime/HostOp changes. C/Worker provider transport remains open.

  • CM-A08 table logical/stdlib slice (ADR-0032): stdlib/data/table.mq.md adds select/filter/sort/append/set-cell/join/group as value-returning .mq.md functions; Value::Table semantics and GFM AST laziness unchanged; no physical storage migration or new table builtin.

  • CM-A07 Compat-A prototype (ADR-0027): stdlib/ provides format.显示, UTF-8 encoding and io.写值 (resource.write(stdout)), and its output is comparison-tested byte-for-byte against builtin 打印 for Unit/None/Bool/Int/Float/Text/List/Map/Table. Builtin 打印 remains the default; no alias routing, bundled stdlib or HostOp::Print removal; the other value types lack public observation.

Changed

  • Assignment targets are no longer restricted to bare identifiers: a single index place is accepted at statement level. Member assignment stays a parse error; chained assignment stays a parse error. spec/expression.ebnf, spec/evaluation.md, spec/value-wire.md, spec/analysis-v2.md and spec/security.md document the new semantics.
  • docs/STABLE_RELEASE_IMPLEMENTATION.md is the single next-phase implementation plan. Older plan documents are historical input, not current status.
  • AST conformance tests fail early when a golden contains CR bytes instead of 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

  • Windows CI AST golden failure caused by CRLF checkout of canonical JSON 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

  • Linux strict is implemented and probe-gated. It is Available only when a real Bubblewrap smoke test succeeds; otherwise it stays Unavailable and refuses. There is no silent downgrade to project.
  • Strict L1 covers pure Marqdo and native Math; Foreign, native/worker plugins, kernels, external grants and trace files are rejected before the 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 8af7262 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)

  • Markdown profile marqdo.md.v1 with 150 AST goldens.
  • Native Rust interpreter, analysis and execution conformance corpus.
  • lock v3 package-instance graph with full transitive closure, verified Git cache and atomic .marqdo/env materialization (ADR-0016).
  • Analysis v2 and Trace v2 with independent Python consumers and JSON Schemas.
  • Tool Protocol v1 with SemVer/protocol negotiation and whole-bundle digests.
  • Plugin ABI v1 native and worker transports with a brokered fs.read capability path (ADR-0018).
  • Pinned Foreign provider/environment digests and one-shot Python/Node kernels.