lutaml-model 0.8.62 → 0.8.64

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 93221695e02a0814183bb7935d242d416e7b5fc98fdae729ae7b7a548276fa1a
-  data.tar.gz: a8c6631206d1ed3dc635012999f11a2deb9f29409a441a4b85e8d47dd7f2fa9c
+  metadata.gz: 8a6392d2dcfd63cc0cb61e585a164c72e09438c25154651cf9ad935e32985429
+  data.tar.gz: 492ad5dd9799652c1dd6abc9e148edba5937f3a613e143bdf556f625d94851fc
 SHA512:
-  metadata.gz: db7ee4ad6d440102cbf7c2c8452b20844c1ddb251a44f86d44f0655e14f359eb1e1fa17debc2f0857485c34b09c9199c6921c31a55b1910aa103ac0fd5e11ff8
-  data.tar.gz: c8a43921457c96865466b5cb0111ccd4e6aba04585bb6e398bc7ad6f808a2009f2bbffa48e8dc1348ba10937776c96c0707249ab54d0bc018d900c4799eb98e3
+  metadata.gz: c14bf78b03a2cce38a13d7a7c09cc07a7cca4b1dbed59a46c0faa9a5c4c275c872dcbff23afae3f8b8518b4fa926ad64597f0f7d67a7567a711ea05d60f8754d
+  data.tar.gz: c8d8b9723545747d826a7c16bb3d11699963299b2a238b5c04e5a724b1de555463fb839689cef9e46e52137c92a1f6831bd50676097653d888a92c57d7255f29

data/lib/lutaml/model/version.rb

@@ -2,6 +2,6 @@
 
 module Lutaml
   module Model
-    VERSION = "0.8.62"
+    VERSION = "0.8.64"
   end
 end

data/lib/lutaml/xml/adapter/xml_parser.rb

@@ -20,6 +20,7 @@ module Lutaml
           raw_xml = xml
           xml = normalize_xml_for_parse(xml)
           parsed = parse_with_moxml(xml, parse_encoding)
+          assert_no_truncated_recovered_parse!(parsed)
           root_element = parsed.root
 
           raise_empty_document_error if root_element.nil?
@@ -30,6 +31,31 @@ module Lutaml
           new(root, parse_encoding, **parse_document_options(raw_xml))
         end
 
+        # Recover-mode parsing must never masquerade input truncation as
+        # success.
+        #
+        # moxml releases before 0.5.84 parsed Nokogiri documents without
+        # XML_PARSE_HUGE, so libxml2 capped its input buffer at 10 MB;
+        # longer documents tripped a fatal "Resource limit exceeded:
+        # Buffer size limit exceeded, try XML_PARSE_HUGE" error mid-input.
+        # In recover mode Nokogiri records that error on the document and
+        # returns the partial tree, so every node past the limit silently
+        # vanishes (lutaml-model#871). A resource-limit fatal means the
+        # engine stopped early with input left — refuse to deserialize the
+        # partial document instead of dropping trailing content without a
+        # trace.
+        #
+        # moxml 0.5.84 and later always pass XML_PARSE_HUGE, so the
+        # Nokogiri path no longer truncates and this guard stays silent;
+        # it remains for older moxml releases and for any other adapter
+        # that reports the resource-limit family.
+        #
+        # Other recover-mode fatals (e.g. an XML declaration after leading
+        # whitespace) do not drop content and keep the long-standing
+        # recover behavior; only the resource-limit family is truncating.
+        TRUNCATION_FATAL_MARKER =
+          /Resource limit exceeded|Buffer size limit exceeded/
+
         private
 
         def normalize_xml_for_parse(xml)
@@ -62,6 +88,23 @@ module Lutaml
           end
         end
 
+        def assert_no_truncated_recovered_parse!(parsed)
+          errors = parsed.parse_errors
+          return if errors.nil? || errors.empty?
+
+          fatal = errors.find do |message|
+            message.match?(TRUNCATION_FATAL_MARKER)
+          end
+          return if fatal.nil?
+
+          raise Lutaml::Model::InvalidFormatError.new(
+            :xml,
+            "the XML engine reported a fatal resource limit and recovered " \
+            "with a truncated document; refusing to deserialize " \
+            "partial input (#{fatal})",
+          )
+        end
+
         def parse_document_options(xml)
           {
             doctype: extract_doctype_from_xml(xml),

data/lutaml-model.gemspec

@@ -38,7 +38,10 @@ Gem::Specification.new do |spec|
   spec.add_dependency "canon"
   spec.add_dependency "concurrent-ruby"
   spec.add_dependency "liquid", ">= 4.0", "< 6.0"
-  spec.add_dependency "moxml", "~> 0.5.56"
+  # 0.5.84 always passes XML_PARSE_HUGE to libxml2: very large documents
+  # parse completely instead of being silently truncated at the 10 MB
+  # input-buffer cap (lutaml-model#871).
+  spec.add_dependency "moxml", "~> 0.5.84"
   spec.add_dependency "ostruct"
   spec.add_dependency "rubyzip", "~> 3.4"
   spec.add_dependency "thor"

data/spec/lutaml/xml/adapter/over_cap_parse_spec.rb

@@ -0,0 +1,105 @@
+# frozen_string_literal: true
+
+require "spec_helper"
+require "nokogiri"
+require "lutaml/model"
+require "lutaml/xml"
+require "lutaml/xml/adapter/nokogiri_adapter"
+
+# Regression coverage for lutaml-model#871: very large documents must
+# deserialize completely, with no silent truncation and no data loss.
+#
+# libxml2's default mode caps its input buffer at 10 MB and, in recover
+# mode, silently drops every node past the cap. moxml 0.5.84 always
+# passes XML_PARSE_HUGE, so the Nokogiri path now parses past the cap.
+# XmlParser additionally refuses recovered parses that report the
+# resource-limit fatal family, which still protects moxml releases that
+# truncate.
+RSpec.describe "parsing documents past libxml2's 10 MB buffer cap" do
+  let(:annex_class) do
+    Class.new(Lutaml::Model::Serializable) do
+      attribute :id, :string
+
+      xml do
+        element "annex"
+        map_attribute "id", to: :id
+      end
+    end
+  end
+
+  let(:root_class) do
+    annex = annex_class
+
+    Class.new(Lutaml::Model::Serializable) do
+      attribute :annexes, annex, collection: true
+
+      xml do
+        element "metanorma"
+        map_element "annex", to: :annexes
+      end
+    end
+  end
+
+  def build_document(chunk_count:, chunk_size:, with_tail: false)
+    filler = "A" * chunk_size
+    document = +"<metanorma>"
+    chunk_count.times do |index|
+      document << %(<annex id="annex#{index}" data="#{filler}"/>)
+    end
+    document << %(<annex id="annex-tail-marker"/>) if with_tail
+    document << "</metanorma>"
+    document
+  end
+
+  # Six 1.7 MB attribute values push the cumulative input past libxml2's
+  # 10 MB buffer cap mid-document, mirroring the issue's failing manifest.
+  def over_cap_document
+    build_document(chunk_count: 6, chunk_size: 1_700_000, with_tail: true)
+  end
+
+  it "confirms libxml2's default mode still trips the buffer cap past 10 MB" do
+    errors = Nokogiri::XML(over_cap_document) { |config| config.recover.nonet }.errors
+
+    expect(errors.map(&:message).join("\n")).to match(/\bFATAL:/)
+  end
+
+  it "parses over-cap documents completely on the nokogiri path" do
+    Lutaml::Model::Config.with_adapter(xml: :nokogiri) do
+      model = root_class.from_xml(over_cap_document)
+
+      expect(model.annexes.map(&:id)).to eq(
+        ["annex0", "annex1", "annex2", "annex3", "annex4", "annex5",
+         "annex-tail-marker"],
+      )
+    end
+  end
+
+  it "returns the complete tree from the adapter layer" do
+    Lutaml::Model::Config.with_adapter(xml: :nokogiri) do
+      document = Lutaml::Xml::Adapter::NokogiriAdapter.parse(over_cap_document)
+
+      expect(document.root.children.count).to eq(7)
+    end
+  end
+
+  it "keeps parsing documents whose fatals do not drop content" do
+    xml = "  <?xml version=\"1.0\"?>\n<metanorma><annex id=\"a\"/></metanorma>"
+
+    Lutaml::Model::Config.with_adapter(xml: :nokogiri) do
+      model = root_class.from_xml(xml)
+
+      expect(model.annexes.map(&:id)).to eq(["a"])
+    end
+  end
+
+  it "keeps parsing smaller documents unchanged" do
+    Lutaml::Model::Config.with_adapter(xml: :nokogiri) do
+      model = root_class.from_xml(
+        build_document(chunk_count: 2, chunk_size: 500_000, with_tail: true),
+      )
+
+      expect(model.annexes.map(&:id))
+        .to eq(["annex0", "annex1", "annex-tail-marker"])
+    end
+  end
+end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: lutaml-model
 version: !ruby/object:Gem::Version
-  version: 0.8.62
+  version: 0.8.64
 platform: ruby
 authors:
 - Ribose Inc.
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-09-25 00:00:00.000000000 Z
+date: 2026-09-26 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: base64
@@ -92,14 +92,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.5.56
+        version: 0.5.84
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.5.56
+        version: 0.5.84
 - !ruby/object:Gem::Dependency
   name: ostruct
   requirement: !ruby/object:Gem::Requirement
@@ -1901,6 +1901,7 @@ files:
 - spec/lutaml/xml/adapter/nokogiri_adapter_spec.rb
 - spec/lutaml/xml/adapter/oga_adapter_spec.rb
 - spec/lutaml/xml/adapter/order_spec.rb
+- spec/lutaml/xml/adapter/over_cap_parse_spec.rb
 - spec/lutaml/xml/adapter/ox_adapter_spec.rb
 - spec/lutaml/xml/adapter/rexml_adapter_spec.rb
 - spec/lutaml/xml/adapter/xml_namespace_spec.rb