[nanvix] E: Remove lxml built-in shim from cpython

.nanvix/build.py

@@ -22,7 +22,7 @@
 
 config = load_sibling("config", __file__)
 docker_mod = load_sibling("docker", __file__)
-lxml_mod = load_sibling("lxml", __file__)
+setup_local_mod = load_sibling("setup_local", __file__)
 
 
 def make_args(
@@ -115,7 +115,7 @@ def build(
         return
     effective_sysroot = config.DOCKER_SYSROOT_PATH if docker else sysroot
     effective_toolchain = config.DOCKER_TOOLCHAIN_PATH if docker else toolchain
-    lxml_mod.generate_setup_local(repo_root, Path(effective_sysroot))
+    setup_local_mod.generate_setup_local(repo_root, Path(effective_sysroot))
     args = make_args(
         effective_sysroot,
         effective_toolchain,

.nanvix/config.py

@@ -341,8 +341,6 @@ def configure_opts(
     # #323 wave 8 — regex and plistlib
     "test_re",
     "test_plistlib",
-    # #600 — lxml built-in smoke test
-    "test_nanvix_lxml",
     # #526 — _lzma stdlib enablement
     "test_lzma",
     # #327 — network and protocol tests (IPv4 only; IPv6 disabled)
@@ -464,9 +462,9 @@ def configure_opts(
     "lib/pkgconfig",
 ]
 
-# site-packages is no longer trimmed because lxml runtime files may be
-# installed there by downstream packaging.  When the directory is empty
-# it remains harmlessly on disk (ramfs.trim_sysroot only removes empty
+# site-packages is not trimmed because downstream third-party packages
+# may install runtime files there.  When the directory is empty it
+# remains harmlessly on disk (ramfs.trim_sysroot only removes empty
 # bin/).  To force-trim site-packages for minimal images, add the path
 # back into SYSROOT_TRIM_DIRS above.
 

.nanvix/docker.py

@@ -286,7 +286,7 @@ def _generate_setup_local_cmd() -> str:
     """Shell command to generate Modules/Setup.local inside the container.
 
     Rendered from .nanvix/setup_local.py (single source of truth shared
-    with the host build path .nanvix/lxml.py::generate_setup_local).
+    with the host build path .nanvix/setup_local.py::generate_setup_local).
     The rendered file body is emitted via a single ``printf '%s\\n' ...``
     invocation with each line single-quoted for the container shell.
     """

.nanvix/nanvix.toml

@@ -15,7 +15,4 @@ sqlite = "3.49.0"
 openssl = "3.5.0"
 bzip2 = "1.0.8"
 libffi = "3.4.6"
-libxml2 = "2.12.9"
-libxslt = "1.1.42"
-lxml = "5.3.0"
 xz = "5.2.5"

.nanvix/package.py

@@ -21,7 +21,6 @@
 
 config = load_sibling("config", __file__)
 build_mod = load_sibling("build", __file__)
-lxml_mod = load_sibling("lxml", __file__)
 ramfs_mod = load_sibling("ramfs", __file__)
 
 
@@ -102,9 +101,6 @@ def package(
     if not sysroot_installed.is_dir():
         raise FileNotFoundError(f"Install did not produce {sysroot_installed}")
 
-    # Stage lxml Python package into the installed sysroot.
-    lxml_mod.stage_lxml_runtime(repo_root, sysroot_installed)
-
     # --- Buildroot: build dependencies ---
     buildroot_pkg = release_staging / "buildroot-pkg"
     buildroot_pkg.mkdir(parents=True)

.nanvix/setup_local.py

@@ -3,9 +3,9 @@
 
 """Single source of truth for Modules/Setup.local on Nanvix builds.
 
-The host-build path (.nanvix/lxml.py::generate_setup_local) and the
-Docker-build path (.nanvix/docker.py::_generate_setup_local_cmd) both
-consume SETUP_LOCAL_ENTRIES and render the same file body via
+The host-build path (generate_setup_local, below) and the Docker-build
+path (.nanvix/docker.py::_generate_setup_local_cmd) both consume
+SETUP_LOCAL_ENTRIES and render the same file body via
 render_setup_local().
 
 Module ordering matters: makesetup applies "first rule wins" semantics
@@ -19,6 +19,7 @@
 from __future__ import annotations
 
 from enum import Enum
+from pathlib import Path
 from typing import Iterable, NamedTuple, Sequence
 
 
@@ -91,31 +92,6 @@ class SetupEntry(NamedTuple):
             "take precedence."
         ),
     ),
-    SetupEntry(
-        name="_lxml_etree",
-        linkage=Linkage.SHARED,
-        tokens=(
-            "lxml_etree_builtin.c",
-            "-L{sysroot}/lib",
-            "-llxml_etree",
-            "-lxslt",
-            "-lexslt",
-            "-lxml2",
-            "-lz",
-        ),
-        section_header=("lxml C extension modules (linked via pre-built archives)."),
-    ),
-    SetupEntry(
-        name="_lxml_elementpath",
-        linkage=Linkage.SHARED,
-        tokens=(
-            "lxml_elementpath_builtin.c",
-            "-L{sysroot}/lib",
-            "-llxml_elementpath",
-            "-lxml2",
-            "-lz",
-        ),
-    ),
     # ---------------- Data primitives (no external deps) ----------------
     *(
         SetupEntry(
@@ -329,9 +305,9 @@ def render_setup_local(
             lines.append(entry.linkage.value)
             current_linkage = entry.linkage
         elif entry.section_header:
-            # Mid-section group separator (e.g. moving from lxml to the
-            # data-primitive modules without changing linkage). Emit as
-            # a blank-line-separated comment block before the entry.
+            # Mid-section group separator (e.g. moving from one shared
+            # group to the next without changing linkage). Emit as a
+            # blank-line-separated comment block before the entry.
             lines.append("")
             lines.extend(_wrap_comment(entry.section_header))
 
@@ -345,3 +321,20 @@ def render_setup_local(
 
     lines.append("")  # trailing newline
     return "\n".join(lines)
+
+
+def generate_setup_local(repo_root: Path, sysroot: Path) -> None:
+    """Generate Modules/Setup.local from SETUP_LOCAL_ENTRIES.
+
+    Single source of truth shared with the Docker build path
+    (.nanvix/docker.py::_generate_setup_local_cmd). The {sysroot}
+    placeholder in entry tokens is substituted with the host build's
+    sysroot path here.
+    """
+    setup_local = repo_root / "Modules" / "Setup.local"
+    content = render_setup_local(
+        sysroot=str(sysroot),
+        header_comment="Auto-generated by .nanvix/setup_local.py -- do not edit manually.",
+    )
+    setup_local.write_text(content, encoding="utf-8")
+    print(f"[setup_local] Generated {setup_local}")

.nanvix/test.py

@@ -31,7 +31,6 @@
 
 config = load_sibling("config", __file__)
 build_mod = load_sibling("build", __file__)
-lxml_mod = load_sibling("lxml", __file__)
 ramfs_mod = load_sibling("ramfs", __file__)
 
 
@@ -547,11 +546,7 @@ def stage(
         )
 
     # Copy test script — a simple smoke test that validates the interpreter.
-    # The lxml import test is only included for standalone mode because
-    # xmlInitParser() hangs in multi-process/single-process modes where
-    # filesystem I/O goes through nanvixd's virtualized host-FS layer.
     hello_script = sysroot_dir / "test_hello.py"
-    standalone = process_mode == "standalone"
     # Phase 0 of the .a -> .so migration: `array` is now a shared
     # extension at lib/python3.12/lib-dynload/array.cpython-312.so
     # (built from `*shared* array arraymodule.c` in Setup.local).
@@ -567,26 +562,12 @@ def stage(
         "print(f'CPYTHON_TEST_ARRAY_SO: array loaded via dlopen from "
         "{array.__file__}')\n"
     )
-    lxml_snippet = (
-        "try:\n"
-        "    import lxml.etree\n"
-        "    doc = lxml.etree.fromstring(b'<root><child>lxml OK</child></root>')\n"
-        "    assert doc.tag == 'root'\n"
-        "    assert doc[0].text == 'lxml OK'\n"
-        "    print('CPYTHON_TEST_LXML: lxml.etree import and parse OK')\n"
-        "except ImportError as e:\n"
-        "    print(f'CPYTHON_TEST_LXML_SKIP: {e}')\n"
-        "except Exception as e:\n"
-        "    print(f'CPYTHON_TEST_LXML_FAIL: {e}')\n"
-        "    sys.exit(1)\n"
-    )
     hello_script.write_text(
         "import sys\n"
         "print('CPYTHON_TEST_HELLO: Hello from Python', sys.version_info[:2])\n"
         "print('CPYTHON_TEST_PLATFORM:', sys.platform)\n"
         + array_snippet
-        + _render_so_sanity_snippets()
-        + (lxml_snippet if standalone else ""),
+        + _render_so_sanity_snippets(),
     )
 
     # Copy the HTTP server smoke-test script from the repo root into the
@@ -823,8 +804,6 @@ def run_hello(
     environment variable syntax.  Multi-process and single-process modes
     use direct host-filesystem access (no ramfs).
     """
-    standalone = process_mode == "standalone"
-
     print(f"Test: Hello world ({process_mode})...")
 
     returncode, output, elapsed_ms = _run_nanvixd_script(
@@ -846,26 +825,18 @@ def run_hello(
 
     # Validate output.
     found_hello = False
-    found_lxml = False
     for line in output.splitlines():
         if line.startswith("CPYTHON_TEST_"):
             tag = line.split(":")[0].replace("CPYTHON_TEST_", "")
             print(f"  {tag}: {line.strip()}")
             if tag == "HELLO":
                 found_hello = True
-            elif tag in ("LXML", "LXML_SKIP"):
-                found_lxml = True
 
     if not found_hello:
         print("  FAIL: Hello test did not produce expected output")
         print(output)
         raise RuntimeError("Hello test did not produce expected output")
 
-    if standalone and not found_lxml:
-        # lxml staging is best-effort — if the runtime package was not
-        # available (e.g. release asset missing), the test is non-fatal.
-        print("  WARNING: lxml import/parse test did not produce expected output")
-
     print("  PASS")
 
 
@@ -1194,7 +1165,6 @@ def run_all(
         run_fn=run_fn,
         docker=docker,
     )
-    lxml_mod.stage_lxml_runtime(repo_root, staging / "sysroot")
 
     # Ramfs — only needed for standalone mode.  Multi-process and
     # single-process use host-filesystem access (no ramfs).