volaro 0.1.0-alpha.6 → 0.1.0-alpha.7

package/README.md

@@ -1,6 +1,6 @@
 # Volaro
 
-**Limited alpha** (this is `0.1.0-alpha.6`), published under both the
+**Limited alpha** (this is `0.1.0-alpha.7`), published under both the
 `latest` and `alpha` dist-tags: `npm install volaro`.
 
 Volaro is an experimental application language intended for AI authoring and

package/bin/vl.js

@@ -69,6 +69,11 @@ function pyInvocation(pp, moduleName, moduleArgs) {
     ...process.env,
     PYTHONPATH: [pp, VALIDATOR_PP, process.env.PYTHONPATH].filter(Boolean).join(delimiter),
     PYTHONDONTWRITEBYTECODE: "1",
+    // Diagnostics and build summaries contain non-ASCII (box drawing, arrows).
+    // Without this, Windows encodes a redirected stdout with the legacy code
+    // page and the compiler dies with UnicodeEncodeError; the capturing path
+    // below also decodes the child's output as UTF-8.
+    PYTHONIOENCODING: "utf-8",
   };
   return { cmd: py.cmd, argv: ["-B", "-m", moduleName, ...moduleArgs], env };
 }

package/compiler/SOURCE_INFO.json

@@ -1,6 +1,6 @@
 {
-  "revision": "78304efecb7e728025c959e1fdf207749b3652ed",
+  "revision": "3469656bab89d0e379c1e31cb19013a21fdef556",
   "source": "git",
   "dirty": false,
-  "content_sha256": "1747ce8d53c4ae2caed129b014f01f7e4af5619ab5358a833e0c888e940c830e"
+  "content_sha256": "656e9ef8c494c763da68868cf0f80ce509bf51bfd9e71b79ae153407eeb16e83"
 }

package/compiler/SOURCE_REV

@@ -1 +1 @@
-78304efecb7e728025c959e1fdf207749b3652ed
+3469656bab89d0e379c1e31cb19013a21fdef556

package/compiler/vlbuild/vlbuild/static_assets.py

@@ -2,7 +2,9 @@
 from __future__ import annotations
 
 import json
+import os
 import shutil
+import stat
 from pathlib import Path
 from urllib.parse import unquote, urlsplit
 
@@ -10,6 +12,57 @@ from vlcheck import ast_nodes as A
 from .emit import BuildError, _walk
 
 
+# A Windows reparse tag with this bit set names another file or directory:
+# symbolic links, directory junctions (mount points) and similar redirects.
+# Cloud-file placeholders (OneDrive) do not set it, so they still count as files.
+_NAME_SURROGATE = 0x20000000
+
+
+def _is_link(path: Path) -> bool:
+    """True for a symbolic link or a Windows junction; neither is followed.
+
+    `Path.is_symlink()` is False for an NTFS junction, and `Path.is_junction()`
+    only exists from Python 3.12, so older Pythons read the reparse tag.
+    """
+    try:
+        st = os.lstat(path)
+    except OSError:
+        return False
+    if stat.S_ISLNK(st.st_mode):
+        return True
+    is_junction = getattr(path, "is_junction", None)
+    if is_junction is not None and is_junction():
+        return True
+    return bool(getattr(st, "st_reparse_tag", 0) & _NAME_SURROGATE)
+
+
+def _asset_files(root: Path) -> list[Path]:
+    """Every regular file under root, refusing links and never descending into one."""
+    files = []
+    pending = [root]
+    while pending:
+        folder = pending.pop()
+        for path in sorted(folder.iterdir()):
+            if _is_link(path):
+                raise BuildError(
+                    f"E-ASSET-LINK: symbolic links and junctions are not supported in assets/: {path}")
+            if path.is_dir():
+                pending.append(path)
+            elif path.is_file():
+                files.append(path)
+            else:
+                raise BuildError(f"E-ASSET-PATH: assets/ supports regular files only: {path}")
+    # Defence in depth: whatever the checks above missed, every file must
+    # still resolve to somewhere inside the resolved assets/ root.
+    real_root = root.resolve()
+    for path in files:
+        real = path.resolve()
+        if real != real_root and real_root not in real.parents:
+            raise BuildError(
+                f"E-ASSET-LINK: {path} resolves outside assets/; symbolic links and junctions are not supported")
+    return sorted(files)
+
+
 def prepare_assets(entry: Path, bundle: Path, emitter) -> tuple[Path, list[Path]]:
     root = entry.parent / "assets"
     dest = bundle / "assets"
@@ -17,23 +70,17 @@ def prepare_assets(entry: Path, bundle: Path, emitter) -> tuple[Path, list[Path]
     if root.exists() and (dest.resolve() == root.resolve() or root.resolve() in dest.resolve().parents
                           or dest.resolve() in root.resolve().parents):
         raise BuildError("E-ASSET-OUTPUT: output assets/ overlaps source assets/; choose another -o directory")
-    if dest.is_symlink():
-        raise BuildError("E-ASSET-OUTPUT: output assets/ must not be a symbolic link")
+    if _is_link(dest):
+        raise BuildError("E-ASSET-OUTPUT: output assets/ must not be a symbolic link or junction")
     if dest.exists() and not dest.is_dir():
         raise BuildError("E-ASSET-OUTPUT: output assets/ must be a directory")
     files = []
-    if root.is_symlink():
-        raise BuildError("E-ASSET-LINK: assets/ must not be a symbolic link")
+    if _is_link(root):
+        raise BuildError("E-ASSET-LINK: assets/ must not be a symbolic link or junction")
     if root.exists():
         if not root.is_dir():
             raise BuildError("E-ASSET-PATH: assets must be a directory beside the entry file")
-        for path in sorted(root.rglob("*")):
-            if path.is_symlink():
-                raise BuildError(f"E-ASSET-LINK: symbolic links are not supported in assets/: {path}")
-            if path.is_file():
-                files.append(path)
-            elif not path.is_dir():
-                raise BuildError(f"E-ASSET-PATH: assets/ supports regular files only: {path}")
+        files = _asset_files(root)
     # Check literal img references into this convention directory. Other URLs
     # and dynamic strings retain their existing meaning; no guessing/rewrite.
     for view in emitter.views:

package/language/crib.md

@@ -748,7 +748,7 @@ No `alt` is `E-IMG-ALT`. Put local files in `assets/` beside the entry file:
 `img src:"assets/logo.svg" alt:"Logo"`. Build and dev copy the directory into
 the served bundle; dev rebuilds on asset additions, edits and deletions.
 A missing literal `assets/...` image is `E-ASSET-MISSING` at build time.
-The output `assets/` directory is compiler-owned; source symlinks are rejected.
+The output `assets/` directory is compiler-owned; source symlinks and junctions are rejected.
 Other relative paths are not copied automatically; `data:` and external URLs work.
 
 Every build that renders a page needs a module-level `page title:"…"` (becomes

package/language/spec.md

@@ -2536,5 +2536,6 @@ hex digits representing a Unicode scalar value; surrogates are invalid.
 The prototype copies the entry file's sibling `assets/` directory into the
 browser bundle at `assets/` for both build and dev. Literal image references
 inside this directory must exist at build time; dynamic URLs are not checked.
-The directory contains public files only, cannot contain symbolic links, and
-is copied without transformations. Output `assets/` is compiler-owned.
+The directory contains public files only, cannot contain symbolic links or
+Windows junctions, and is copied without transformations. Output `assets/` is
+compiler-owned.

package/language/supported.md

@@ -482,8 +482,8 @@ Dev uses the same build path and watches asset additions, edits and deletions.
 Refresh the browser after a rebuild; this does not add automatic browser reload.
 
 Literal image URLs into `assets/` are checked before output is written.
-Missing files fail with `E-ASSET-MISSING`; traversal paths and symbolic links
-are rejected. Query strings and URL-encoded filenames are supported. Dynamic
+Missing files fail with `E-ASSET-MISSING`; traversal paths, symbolic links and
+Windows junctions are rejected (`E-ASSET-LINK`). Query strings and URL-encoded filenames are supported. Dynamic
 URLs are not existence-checked, but all files in the directory are copied.
 All regular files in `assets/`, including dotfiles, are public: put only intended
 public resources there. Output `assets/` is compiler-owned and replaced on each

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "volaro",
-  "version": "0.1.0-alpha.6",
+  "version": "0.1.0-alpha.7",
   "description": "Volaro \u2014 an application language written to be authored by an AI agent and read by a person. Ships the language reference and a working compiler for a supported subset (volaro check / build / dev). CLI: `volaro` (alias `vl`).",
   "repository": {
     "type": "git",

package/scripts/selftest.mjs

@@ -48,6 +48,16 @@ try {
   const badChk = spawnSync(node, [vl, "check", bad], { encoding: "utf8" });
   check("vl check exits non-zero on broken source", badChk.status !== 0);
 
+  // A diagnostic's source frame uses box-drawing characters. Windows gives a
+  // redirected Python stdout a legacy code page (cp1252), which cannot
+  // encode them; PYTHONIOENCODING=ascii reproduces that on every OS.
+  const narrow = spawnSync(node, [vl, "check", bad],
+    { encoding: "utf8", env: { ...process.env, PYTHONIOENCODING: "ascii" } });
+  const narrowOut = (narrow.stdout || "") + (narrow.stderr || "");
+  check("vl check prints its diagnostic through a non-UTF-8 stdout",
+    !narrowOut.includes("UnicodeEncodeError") && narrowOut.includes("error[E-") && narrowOut.includes("┌─"),
+    narrowOut.trim().split("\n").pop());
+
   // `volaro routes <project>`: page-server route manifest inspection
   const posts = join(dir, "app", "posts");
   mkdirSync(posts, { recursive: true });