moxml 0.5.26 → 0.5.27

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: b7d4ecc2bb794f60a6b240faf3828c7847da9eebf8d45e44b35dd863f8ef7b09
-  data.tar.gz: 04d2ecbc979863a35825a8ba95a7897d4f7d395629c5ca959dfa8fd96c71a662
+  metadata.gz: b9d156e4f8dfd71fc334b5626690f325a78abf95496bf19b6141bf649beeee58
+  data.tar.gz: 5f3c24b3211e8381d14a34c7c23781a808e2f55b1d73d06af3be0fe7d3640a75
 SHA512:
-  metadata.gz: a63b4c25f0838edbb6ba6cb787b68000e614afbfbb700a3167a5222ab7219fef4f7add25a009991de44b78eeb2b117c96f0743655e0fa058e4ac3d697877093b
-  data.tar.gz: c8700770e6873e0a9af730cc94b149c019d63df0aadd585f3f382e1c4fe4246d9d4bb243e9f8347ed4bdb83b67d84aa497de69152cf3ea2921c6be55b9037d05
+  metadata.gz: 1c5cd5d01ece3fcced5049148d2df776a1dcb9b1e95b0fe9a145caef05c0e2d5080b32214c7eb9f8f7bb277412827b413b3c59e38a49162241208944791a6ed6
+  data.tar.gz: 33faf5c9269c3c74a6def61ef3933a9bdc2caedf484e0144ca1c8e6465aeb1640bb47c3f44ce9c777ed5bbe8e9682efd840cb87d4eceedac764f4c496519fff3

data/lib/moxml/adapter/leptris/serialize.rb

@@ -85,26 +85,37 @@ module Moxml
         # detection below is a no-op on correct builds.
         RAW_AMP_RE = /&(?!#{Entity::NAME_PATTERN};|#\d+;|#x[0-9A-Fa-f]+;)/
 
+        # A raw "<" in text position — the engine intermittently
+        # loses the escape when parsing under allocation pressure
+        # (leptris-ruby#131), so a decoded &#x3c; serializes bare and
+        # reparsing truncates at it. Valid markup NEVER puts "<"
+        # before a non-name, non-delimiter character, so this is
+        # deterministic in markup segments (CDATA is literal and
+        # excluded by the segment walker).
+        RAW_LT_RE = /<(?![A-Za-z_:\/?!])/
+
         def normalize_serialization(xml, options)
           # The libxml2-layout serializer (>= 1.9.42) keeps attribute
           # apostrophes literal; older engines escaped them.
           needs_apos = !LIBXML2_LAYOUT_PARITY && xml.include?("&apos;")
           needs_expand = options[:expand_empty] && xml.include?("/>")
-          # One scan: the bare-ampersand regex fails fast on
-          # ampersand-free output — no include? pre-filter needed.
+          # One scan each: both regexes fail fast on clean output.
           needs_amp = xml.match?(RAW_AMP_RE)
-          return xml unless needs_apos || needs_expand || needs_amp
+          needs_lt = xml.match?(RAW_LT_RE)
+          return xml unless needs_apos || needs_expand || needs_amp || needs_lt
 
           out = +""
           pos = 0
           while pos < xml.length
             opener_at, terminator = next_literal_region(xml, pos)
             if opener_at.nil?
-              out << normalize_markup(xml[pos..], needs_apos, needs_expand, needs_amp: needs_amp)
+              out << normalize_markup(xml[pos..], needs_apos, needs_expand,
+                                      needs_amp: needs_amp, needs_lt: needs_lt)
               break
             end
 
-            out << normalize_markup(xml[pos...opener_at], needs_apos, needs_expand, needs_amp: needs_amp)
+            out << normalize_markup(xml[pos...opener_at], needs_apos, needs_expand,
+                                    needs_amp: needs_amp, needs_lt: needs_lt)
             search_from = opener_at + opener_at_offset(terminator)
             close = xml.index(terminator, search_from)
             close_end = close.nil? ? xml.length : close + terminator.length
@@ -136,7 +147,7 @@ module Moxml
           terminator == "-->" ? 4 : 0
         end
 
-        def normalize_markup(markup, needs_apos, needs_expand, needs_amp: false)
+        def normalize_markup(markup, needs_apos, needs_expand, needs_amp: false, needs_lt: false)
           markup = markup.gsub("&apos;", "'") if needs_apos
           if needs_expand
             markup = markup.gsub(EMPTY_ELEMENT_RE) do
@@ -144,8 +155,9 @@ module Moxml
             end
           end
           # Segment-aware: this never sees CDATA/comment/PI content,
-          # where a bare & is literal data.
+          # where bare & and < are literal data.
           markup = markup.gsub(RAW_AMP_RE, "&amp;") if needs_amp
+          markup = markup.gsub(RAW_LT_RE, "&lt;") if needs_lt
           markup
         end
       end

data/lib/moxml/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Moxml
-  VERSION = "0.5.26"
+  VERSION = "0.5.27"
 end

data/spec/moxml/adapter/leptris_spec.rb

@@ -243,6 +243,36 @@ RSpec.describe Moxml::Adapter::Leptris do
       expect(doc.root.children.to_a.select(&:text?)).to be_empty
     end
 
+    it "repairs a raw < in text position (issue #167 / leptris-ruby#131)" do
+      # The engine intermittently drops the escape when parsing under
+      # allocation pressure: a decoded &#x3c; serializes bare and the
+      # output fails to reparse. Valid markup never puts < before a
+      # non-name character, so the damage is deterministically
+      # repairable in markup segments.
+      repaired = described_class.normalize_serialization(
+        %(<?xml version="1.0"?><r><pre>A <\n B</pre></r>\n), {}
+      )
+      expect(repaired).to include("A &lt;")
+      expect(repaired).not_to match("A <")
+    end
+
+    it "keeps a literal < inside CDATA untouched by the repair" do
+      doc = ctx.parse(%(<r><![CDATA[A < B]]></r>))
+      expect(doc.to_xml).to include("<![CDATA[A < B]]>")
+    end
+
+    it "survives parse-under-allocation-pressure without emitting raw < (issue #167)" do
+      xml = "<?xml version=\"1.0\"?><doc>#{%(<s><pre alt="A B">A &#x3c;\n B</pre></s>) * 10}</doc>"
+      30.times do
+        100.times { |j| "pressure#{j}" * 20 }
+        out = ctx.parse(xml, noblanks: true).to_xml(
+          declaration: true, encoding: "UTF-8", indent: 2, expand_empty: false,
+        )
+        expect(out).not_to match("A <"), "raw unescaped < leaked into text"
+        expect(Nokogiri::XML(out, &:strict)).to be_a(Nokogiri::XML::Document)
+      end
+    end
+
     it "refuses noblanks on readonly parses while the strip path is active" do
       # The strip mutates the tree; readonly freezes it at parse. On
       # bindings whose engine flag is libxml2-safe (probe), the flag

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: moxml
 version: !ruby/object:Gem::Version
-  version: 0.5.26
+  version: 0.5.27
 platform: ruby
 authors:
 - Ribose Inc.