makiri 0.2.0-aarch64-linux → 0.3.0-aarch64-linux

data/script/build_native_gem.rb

@@ -24,7 +24,7 @@ root = File.expand_path("..", __dir__)
 spec = Gem::Specification.load(File.join(root, "makiri.gemspec"))
 
 libs = Dir[File.join(root, "lib", "makiri", "*", "makiri.{so,bundle}")].sort
-abort "no precompiled libraries found under lib/makiri/*/ — stage them first" if libs.empty?
+abort "no precompiled libraries found under lib/makiri/*/ - stage them first" if libs.empty?
 
 # Native gem: ship binaries, not the C sources or the vendored Lexbor tree, and
 # declare no extension so install never tries to compile.
@@ -34,7 +34,7 @@ spec.files = spec.files.reject { |f| f.start_with?("ext/", "vendor/") }
 spec.files += libs.map { |p| p.sub("#{root}/", "") }
 spec.files.uniq!
 
-# Bound the Ruby versions this binary gem serves — one subdir per ABI minor.
+# Bound the Ruby versions this binary gem serves - one subdir per ABI minor.
 abis = libs.map { |p| File.basename(File.dirname(p)) }.sort_by { |v| v.split(".").map(&:to_i) }
 lo_major, lo_minor = abis.first.split(".").map(&:to_i)
 hi_major, hi_minor = abis.last.split(".").map(&:to_i)

data/script/check_c_safety.rb

@@ -62,6 +62,38 @@ RULES = [
     message: "mkr_verified_text_t must be minted only by mkr_verified_text_from_view (the validated boundary)",
     regex: /\(\s*mkr_verified_text_t\s*\)\s*\{/
   ),
+  # --- HTML/XML representation boundary (see docs/html_xml_boundary_hardening) ---
+  # These symbols assume one DOM representation; using them outside their
+  # representation-correct / kind-checked home is how shared glue (XPath, NodeSet,
+  # node identity) silently treats an XML node as HTML (or vice versa) - an
+  # assert-abort or memory type-confusion. Each is allowlisted only in the files
+  # that legitimately own it; anywhere else trips the lint.
+  Rule.new(
+    id: "html_doc_unwrap_boundary",
+    message: "mkr_html_doc_unwrap is HTML-only; shared/XML code must use the kind-aware mkr_node_unwrap",
+    regex: /\bmkr_html_doc_unwrap\s*\(/
+  ),
+  Rule.new(
+    id: "parsed_html_doc_boundary",
+    message: "mkr_parsed_html_doc (asserts kind==HTML) may only be used in a kind-checked / HTML-only site",
+    regex: /\bmkr_parsed_html_doc\s*\(/
+  ),
+  Rule.new(
+    id: "parsed_xml_doc_boundary",
+    message: "mkr_parsed_xml_doc may only be used in a kind-checked / XML-representation site",
+    regex: /\bmkr_parsed_xml_doc\s*\(/
+  ),
+  Rule.new(
+    id: "owner_document_boundary",
+    message: "owner_document is an HTML-only lxb field; shared code must compare documents via mkr_node_document",
+    regex: /\bowner_document\b/
+  ),
+  Rule.new(
+    id: "node_raw_boundary",
+    message: "mkr_node_raw is the kind-agnostic raw pointer (identity / kind-guaranteed only); " \
+             "to dereference a node use mkr_html_node or mkr_xml_node_unwrap (kind-checked)",
+    regex: /\bmkr_node_raw\s*\(/
+  ),
 ].freeze
 
 def load_config

data/script/check_c_safety_allowlist.yml

@@ -10,3 +10,86 @@ ignore_paths:
   - path: ext/makiri/bridge/text_token.c
     rule: verified_text_forge
     reason: mkr_verified_text_from_view, the sole sanctioned mint of mkr_verified_text_t; its input view is already validated by the bridge string helpers.
+  # --- HTML/XML representation boundary: each per-rep symbol is exempt only in
+  #     the files that own it (declaration, definition, or a kind-checked use), so
+  #     a new use anywhere else — especially shared glue — trips the lint ---
+  # mkr_html_doc_unwrap (HTML-only document unwrap)
+  - path: ext/makiri/glue/glue.h
+    rule: html_doc_unwrap_boundary
+    reason: declaration of the HTML-only document unwrap.
+  - path: ext/makiri/glue/ruby_doc.c
+    rule: html_doc_unwrap_boundary
+    reason: definition + HTML Document methods (parse/serialize/title/compat_mode) — all HTML-only.
+  - path: ext/makiri/glue/ruby_node.c
+    rule: html_doc_unwrap_boundary
+    reason: the kind-aware mkr_node_raw calls it only on its HTML branch (after a MKR_DOC_XML check).
+  - path: ext/makiri/glue/ruby_html_node.c
+    rule: html_doc_unwrap_boundary
+    reason: mkr_html_node_unwrap resolves an HTML Document (after rejecting an XML one at the type boundary).
+  - path: ext/makiri/glue/ruby_html_mutate.c
+    rule: html_doc_unwrap_boundary
+    reason: HTML tree/fragment mutation; XML mutation has its own arena path.
+  - path: ext/makiri/glue/ruby_xpath.c
+    rule: html_doc_unwrap_boundary
+    reason: the HTML branch of mkr_xpath_context_for, entered only after the XML kind returns early.
+  # mkr_parsed_html_doc (asserts kind == HTML)
+  - path: ext/makiri/lexbor_compat/compat.h
+    rule: parsed_html_doc_boundary
+    reason: declaration of the HTML parsed-document accessor.
+  - path: ext/makiri/lexbor_compat/post_parse.c
+    rule: parsed_html_doc_boundary
+    reason: definition (the kind assert lives here) + HTML post-parse pipeline.
+  - path: ext/makiri/glue/ruby_doc.c
+    rule: parsed_html_doc_boundary
+    reason: mkr_html_doc_unwrap is defined here over the HTML parsed document.
+  # mkr_parsed_xml_doc (XML arena accessor) — kept out of the pure-HTML glue files
+  - path: ext/makiri/lexbor_compat/compat.h
+    rule: parsed_xml_doc_boundary
+    reason: declaration of the XML parsed-document accessor.
+  - path: ext/makiri/lexbor_compat/post_parse.c
+    rule: parsed_xml_doc_boundary
+    reason: definition + XML document wrapping/teardown.
+  - path: ext/makiri/glue/ruby_xml.c
+    rule: parsed_xml_doc_boundary
+    reason: XML parse entry / Document construction.
+  - path: ext/makiri/glue/ruby_xml_node.c
+    rule: parsed_xml_doc_boundary
+    reason: XML node read + mutation surface over the arena.
+  - path: ext/makiri/glue/ruby_doc.c
+    rule: parsed_xml_doc_boundary
+    reason: shared Document glue, used only on kind-checked XML branches.
+  - path: ext/makiri/glue/ruby_xpath.c
+    rule: parsed_xml_doc_boundary
+    reason: the XML branch of mkr_xpath_context_for (kind-checked).
+  - path: ext/makiri/glue/ruby_node.c
+    rule: parsed_xml_doc_boundary
+    reason: the kind-aware mkr_node_raw resolves an XML Document only after a MKR_DOC_XML check.
+  # owner_document (HTML-only lxb_dom_node_t field)
+  - path: ext/makiri/lexbor_compat/post_parse.c
+    rule: owner_document_boundary
+    reason: mkr_lxb_document_bytes resolves a Lexbor node's owner document to size its mraw pools (HTML-only; the XML serializer uses arena_bytes instead).
+  - path: ext/makiri/glue/ruby_html_node.c
+    rule: owner_document_boundary
+    reason: HTML node readers operate on lxb_dom_node_t.
+  - path: ext/makiri/glue/ruby_doc.c
+    rule: owner_document_boundary
+    reason: HTML Document operations.
+  - path: ext/makiri/glue/ruby_html_mutate.c
+    rule: owner_document_boundary
+    reason: HTML tree mutation over Lexbor nodes.
+  - path: ext/makiri/xpath/mkr_xpath_node_access_html.h
+    rule: owner_document_boundary
+    reason: the HTML monomorphization of the engine's node-access layer.
+  # mkr_node_raw (kind-agnostic void* raw pointer; never dereferenced as a typed node)
+  - path: ext/makiri/glue/glue.h
+    rule: node_raw_boundary
+    reason: declaration of the kind-agnostic accessor.
+  - path: ext/makiri/glue/ruby_node.c
+    rule: node_raw_boundary
+    reason: defines mkr_node_raw / mkr_node_id (node identity — the pointer is only compared, never dereferenced).
+  - path: ext/makiri/glue/ruby_xpath.c
+    rule: node_raw_boundary
+    reason: the XPath context node / handler-result node, where same-document (hence the representation) is verified before the engine takes the raw pointer.
+  - path: ext/makiri/glue/ruby_node_set.c
+    rule: node_raw_boundary
+    reason: NodeSet.new stores representation-opaque pointers; mkr_node_raw takes a seed node's pointer only after same-document validation guarantees its kind matches the set.

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: makiri
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.3.0
 platform: aarch64-linux
 authors:
 - takahashim
@@ -77,19 +77,26 @@ files:
 - lib/makiri/3.3/makiri.so
 - lib/makiri/3.4/makiri.so
 - lib/makiri/4.0/makiri.so
-- lib/makiri/attribute.rb
-- lib/makiri/cdata.rb
+- lib/makiri/attr.rb
+- lib/makiri/cdata_section.rb
 - lib/makiri/comment.rb
+- lib/makiri/compat_aliases.rb
 - lib/makiri/css.rb
 - lib/makiri/document.rb
 - lib/makiri/document_fragment.rb
 - lib/makiri/document_type.rb
 - lib/makiri/element.rb
+- lib/makiri/html.rb
+- lib/makiri/html/document.rb
+- lib/makiri/html/node_methods.rb
 - lib/makiri/node.rb
 - lib/makiri/node_set.rb
 - lib/makiri/processing_instruction.rb
 - lib/makiri/text.rb
 - lib/makiri/version.rb
+- lib/makiri/xml.rb
+- lib/makiri/xml/document.rb
+- lib/makiri/xml/node_methods.rb
 - lib/makiri/xpath.rb
 - lib/makiri/xpath_context.rb
 - script/build_native_gem.rb

data/lib/makiri/cdata.rb

@@ -1,6 +0,0 @@
-# frozen_string_literal: true
-
-module Makiri
-  class CData < Node
-  end
-end