mcpplibs.mwrap.md4c 0.5.3: MD4C as C++23 modules (md4c-m)

.agents/docs/2026-10-27-add-md4c-mwrap-plan.md

@@ -0,0 +1,55 @@
+# Add mcpplibs.mwrap.md4c (md4c-m 0.5.3)
+
+## Form
+
+- Source: third-party upstream [mity/md4c](https://github.com/mity/md4c), latest tag `release-0.5.3`
+  (`git ls-remote --tags`), MIT. The tarball (`353c346f…059b9e`) is the one compat.md4c 0.5.3 already pins.
+- Form A external mwrap module repository: [mcpplibs/md4c-m](https://github.com/mcpplibs/md4c-m), package
+  `mcpplibs.mwrap.md4c`, version `0.5.3` (the upstream version, first release of the identity, no revision
+  suffix), tag series `mwrap-${version}`.
+- Implementation ownership: the parser `md4c.c` is compat.md4c's (its configuration and ABI are sufficient).
+  Upstream ships a second library, md4c-html (`md4c-html.c`, `entity.c`), which compat.md4c does not build;
+  md4c-m vendors those four files byte for byte under `upstream/md4c/src` (checksums in
+  `upstream/documents.json`). No implementation is linked twice.
+
+## Modules
+
+- `mcpplibs.mwrap.md4c` — every typedef, enum, enumerator, struct and `md_parse` of `<md4c.h>` by
+  `export using`, re-exporting the two below.
+- `mcpplibs.mwrap.md4c.html` — `md_html`.
+- `mcpplibs.mwrap.md4c.macros` — the 23 object-like `MD_FLAG_*` / `MD_DIALECT_*` / `MD_HTML_FLAG_*` macros as
+  `inline constexpr unsigned` in `mcpplibs::mwrap::md4c::macros`. md4c has no function-like or control-flow
+  macros, so there is no side header; a TU that also includes `<md4c.h>` must set the macros aside to name the
+  mapped constants (tested in `tests/header_mix_test.cpp`).
+- `tools/check_exports.py` parses both headers and fails when a public name or flag macro is not exported
+  (negative-checked by removing one export).
+
+## Verification
+
+md4c-m CI (Linux GCC 16.1 and LLVM 23.1.3, macOS LLVM, Windows LLVM, all required): build, gtest suites
+(parser callbacks and details, renderer, header mixing, macros), the export audit, and upstream's own
+`test/run-testsuite.py` over all 11 suites through the `md2html` example (CommonMark 0.31.2 spec 652/652,
+extensions, regressions and coverage: 808 examples). Windows needed binary standard streams in md2html (text
+mode rewrote line endings). The templates job scaffolds the default `render` template as an installed
+package (`tools/test_template.py`) and runs it.
+
+Here, with the CI-pinned mcpp 2026.10.26.1 and `MCPP_INDEX_MIRROR=GLOBAL`:
+
+- `bash tests/run_members.sh md4c-render` — the real release archive through the candidate index; the
+  consumer parses with callbacks (heading level, table cells, task item) and renders GFM HTML:
+  `test result ok`.
+- `python3 tests/mwrap/check_installed_template.py --package mcpplibs.mwrap.md4c@0.5.3` — `mcpp new` from the
+  indexed archive, build and run of the unmodified template: "Verified real indexed archive installation and
+  unmodified template consumer".
+- `mcpp xpkg parse pkgs/mwrap/md4c.lua` (form A, three platforms) and the lint job's Lua checks.
+
+## Mirror and updates
+
+No `mcpp-res` access in this change, so the descriptor uses the GitHub release URL as a plain string (CN
+users fall back to it); a maintainer can add the GitCode mirror with identical bytes. `extensions.autoupdate`
+follows the repository's own `mwrap-${version}` tags.
+
+## Motivation
+
+openmui's Markdown component (openmui-markdown → openmui-ui-markdown) parses with md4c through this module:
+its document model is checked byte for byte against `md_html` on every upstream example.

.github/workflows/validate.yml

@@ -1603,6 +1603,11 @@ jobs:
         shell: bash
         run: python3 tests/mwrap/check_installed_template.py --mcpp "$MCPP" --package mcpplibs.mwrap.imgui@1.92.8
 
+      - name: Installed md4c wrapper archive and actual default render template
+        if: contains(format(' {0} ', env.MEMBERS), ' md4c-render ') && matrix.toolchain == 'default'
+        shell: bash
+        run: python3 tests/mwrap/check_installed_template.py --mcpp "$MCPP" --package mcpplibs.mwrap.md4c@0.5.3
+
       - name: Installed Godot wrapper archive and actual standalone math template
         if: contains(format(' {0} ', env.MEMBERS), ' godot-math ') && matrix.toolchain == 'default'
         shell: bash

mcpp.toml

@@ -16,6 +16,7 @@ members = [
     "tests/mwrap/llamacpp-cpu",
     "tests/mwrap/ffmpeg-cpu",
     "tests/mwrap/imgui-headless",
+    "tests/mwrap/md4c-render",
     "tests/mwrap/grpc-cpu",
     "tests/mwrap/grpc-tools",
     "tests/mwrap/godot-math",

pkgs/mwrap/md4c.lua

@@ -0,0 +1,38 @@
+-- Immutable upstream-aligned mwrap source release (md4c-m mwrap-0.5.3).
+-- The CN mirror is added by a maintainer with mcpp-res access; until then CN
+-- users fall back to the GitHub release (docs/cn-mirror.md).
+package = {
+    spec = "1",
+    namespace = "mcpplibs.mwrap",
+    name = "md4c",
+    description = "md4c-m: MD4C (CommonMark / GFM Markdown parser and md4c-html renderer) as C++23 modules",
+    licenses = { "Apache-2.0" },
+    repo = "https://github.com/mcpplibs/md4c-m",
+    type = "package",
+    extensions = {
+        autoupdate = {
+            tag = "mwrap-${version}",
+            url = "https://github.com/mcpplibs/md4c-m/releases/download/${tag}/md4c-m-${version}.tar.gz",
+        },
+    },
+    xpm = {
+        linux = {
+            ["0.5.3"] = {
+                url = "https://github.com/mcpplibs/md4c-m/releases/download/mwrap-0.5.3/md4c-m-0.5.3.tar.gz",
+                sha256 = "10f7cdffab919bc97e2cf713c0a784e698568eca8a2331ab386515e5bf7743ac",
+            },
+        },
+        macosx = {
+            ["0.5.3"] = {
+                url = "https://github.com/mcpplibs/md4c-m/releases/download/mwrap-0.5.3/md4c-m-0.5.3.tar.gz",
+                sha256 = "10f7cdffab919bc97e2cf713c0a784e698568eca8a2331ab386515e5bf7743ac",
+            },
+        },
+        windows = {
+            ["0.5.3"] = {
+                url = "https://github.com/mcpplibs/md4c-m/releases/download/mwrap-0.5.3/md4c-m-0.5.3.tar.gz",
+                sha256 = "10f7cdffab919bc97e2cf713c0a784e698568eca8a2331ab386515e5bf7743ac",
+            },
+        },
+    },
+}

tests/mwrap/md4c-render/mcpp.toml

@@ -0,0 +1,11 @@
+[package]
+name = "md4c-render"
+version = "0.5.3"
+description = "Consume the immutable md4c wrapper: parse with callbacks and render GFM to HTML"
+
+[indices]
+# Use the complete candidate built-in index prepared by tests/run_members.sh.
+# An explicit empty table overrides the legacy workspace redirect.
+
+[dependencies]
+"mcpplibs.mwrap.md4c" = "0.5.3"

tests/mwrap/md4c-render/tests/behavior.cpp

@@ -0,0 +1,42 @@
+import std;
+import mcpplibs.mwrap.md4c;
+
+namespace macros = mcpplibs::mwrap::md4c::macros;
+
+int main() {
+    const std::string_view md = "# Title\n\n- [x] done\n\n| a | b |\n|:-|-:|\n| 1 | 2 |\n\n~~old~~ www.example.com ©\n";
+
+    // md_html (md4c-html, vendored by the wrapper) with GitHub's dialect.
+    std::string html;
+    if (md_html(md.data(), static_cast<MD_SIZE>(md.size()),
+                [](const MD_CHAR* s, MD_SIZE n, void* out) { static_cast<std::string*>(out)->append(s, n); }, &html,
+                macros::MD_DIALECT_GITHUB, 0) != 0)
+        return 1;
+    for (std::string_view want : {"<h1>Title</h1>", "checked", "<th align=\"left\">a</th>", "<td align=\"right\">2</td>", "<del>old</del>",
+                                  "<a href=\"http://www.example.com\">", "\xC2\xA9"})
+        if (!html.contains(want)) return 2;
+
+    // md_parse (compat.md4c's parser) with callbacks and details.
+    struct Seen {
+        unsigned heading = 0, cells = 0;
+        int      task = 0;
+    } seen;
+    MD_PARSER p{};
+    p.flags = macros::MD_DIALECT_GITHUB;
+    p.enter_block = [](MD_BLOCKTYPE t, void* d, void* u) {
+        auto& s = *static_cast<Seen*>(u);
+        if (t == MD_BLOCK_H) s.heading = static_cast<MD_BLOCK_H_DETAIL*>(d)->level;
+        if (t == MD_BLOCK_TD || t == MD_BLOCK_TH) ++s.cells;
+        if (t == MD_BLOCK_LI) s.task = static_cast<MD_BLOCK_LI_DETAIL*>(d)->is_task;
+        return 0;
+    };
+    p.leave_block = [](MD_BLOCKTYPE, void*, void*) { return 0; };
+    p.enter_span = [](MD_SPANTYPE, void*, void*) { return 0; };
+    p.leave_span = [](MD_SPANTYPE, void*, void*) { return 0; };
+    p.text = [](MD_TEXTTYPE, const MD_CHAR*, MD_SIZE, void*) { return 0; };
+    if (md_parse(md.data(), static_cast<MD_SIZE>(md.size()), &p, &seen) != 0) return 3;
+    if (seen.heading != 1 || seen.cells != 4 || seen.task != 1) return 4;
+
+    std::println("md4c 0.5.3 through import mcpplibs.mwrap.md4c: {} bytes of HTML, {} table cells", html.size(), seen.cells);
+    return 0;
+}