canon 0.3.46 → 0.3.47

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 2dc57b150c1a0a2b9179f4b88a78c22aeb1529aa74d30857d8ca71a6159d6639
-  data.tar.gz: 65db8a7cadb8718b11a475858d3bfd5d4adcf0a4c738a7c940a6dfa60a6bfd3e
+  metadata.gz: 8180f10b0965070c408985ca44868fc14ae2256f24dac607492470aea7e5d076
+  data.tar.gz: e50498b713b972237a67af3b3f5c2cd5fb0cf1cbc3fef0dbb57f628810ce2341
 SHA512:
-  metadata.gz: d411dc92e7978a904f303a95397f301286e9a12a3fc3732bb91919c2564375221b12cb13db64198f41942edeb5bbe2dfc8512b173cbfbdab44d27409d5a009a9
-  data.tar.gz: eaa339eae29ca147aa6ee9a5cecf9a328e69fc9b1f64b3acb742d11f4d451bef89efc11cde012e06e15f21b10ec09f95fbf36e0456c718ba925c015e47dba0ce
+  metadata.gz: 52c5eb3fcb8ab3e09c7de5b8b7120f2baad191467aeccab6154254cd0007128f76ae2f5dbd26102abf19c3cf9f289f3c0445db88917562d7b7edafe9aef7ae5b
+  data.tar.gz: 373032bcba62f9685c41f09bb0d3dce294c50fa8785252e3737231a67f9a0fdfdb5372c85f3b28da5983faf39d4cdf89723dfae8397969511d2688afecaf3418

data/lib/canon/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Canon
-  VERSION = "0.3.46"
+  VERSION = "0.3.47"
 end

data/lib/canon/xml/c14n.rb

@@ -23,19 +23,24 @@ module Canon
       end
 
       # leptris' C-side C14N 1.1 — 23x faster than the Ruby processor
-      # through canon's own API (1MB document). libleptris 1.9.164
-      # closed four of the five leptris#1015 families (attribute
-      # ordering, prefix loss and rebinding, `>` and TAB escaping —
-      # unpinned in the parity spec since gem 1.9.174.0). A local
-      # default-flip attempt surfaced three MORE divergences the
-      # edge corpus did not cover, so the lane stays opt-in
-      # (CANON_C14N_BACKEND=leptris) until they close (leptris#1096):
-      # redundant namespace redeclarations are kept (Ruby removes
-      # them), xmlns:xml with the standard URI is kept (Ruby omits
-      # it), and document-level processing instructions are dropped.
-      # The Ruby lane additionally validates relative namespace URIs
-      # at parse time — a future flip must carry that check into the
-      # native wrapper. Comments mode keeps the Ruby path regardless.
+      # through canon's own API (1MB document) and byte-identical on
+      # the parity corpus, but OPT-IN (CANON_C14N_BACKEND=leptris)
+      # pending three families verified against libxml2 ground truth
+      # (leptris#1096 follow-up): (1) whitespace-only text between
+      # elements is preserved — the Ruby lane's parse drops it and
+      # compact bytes are canon's product for every pretty-printed
+      # document; (2) whitespace-only processing-instruction data is
+      # retained (libxml2 drops it); (3) document-level nodes carry
+      # no "\n" separators (the Ruby lane, like libxml2, inserts
+      # them). libleptris 1.9.176/177 (gem 1.9.177.0) closed the
+      # earlier families — attribute ordering, prefix loss and
+      # rebinding, `>`/TAB escaping, redundant namespace
+      # redeclarations, xmlns:xml omission — and document-level PIs
+      # now serialize in document order (the Ruby processor's
+      # root-first order is non-conformant for prolog PIs).
+      # Relative namespace URIs raise exactly as the Ruby lane does.
+      # Comments mode keeps the Ruby path (the native seam exposes no
+      # with-comments form).
       def self.native_canonicalize(xml, with_comments)
         return nil if with_comments
         return nil if RUBY_ENGINE == "opal"
@@ -44,6 +49,8 @@ module Canon
           Canon::XmlParsing.moxml_adapter_name == :leptris
         return nil unless native_c14n_available?
 
+        validate_relative_namespaces!(xml)
+
         doc = Canon::XmlParsing.moxml_context.parse(xml, readonly: true,
                                                          strict: false)
         begin
@@ -58,6 +65,25 @@ module Canon
         nil
       end
 
+      # Canon::Error parity with the Ruby lane, whose parse rejects
+      # relative namespace URIs. A declaration-shaped scan; a
+      # relative-URI-shaped string inside CDATA is the only false
+      # positive this can produce.
+      RELATIVE_NS_DECL = /xmlns(?::[\w.-]+)?="([^"]*)"/
+      URI_SCHEME = %r{\A[a-zA-Z][a-zA-Z0-9+.-]*:}
+
+      def self.validate_relative_namespaces!(xml)
+        return unless xml.is_a?(String)
+
+        xml.scan(RELATIVE_NS_DECL) do |(uri)|
+          next if uri.nil? || uri.empty?
+
+          unless uri.match?(URI_SCHEME)
+            raise Canon::Error, "Relative namespace URI not allowed: #{uri}"
+          end
+        end
+      end
+
       def self.native_c14n_available?
         return @native_c14n_available unless @native_c14n_available.nil?
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: canon
 version: !ruby/object:Gem::Version
-  version: 0.3.46
+  version: 0.3.47
 platform: ruby
 authors:
 - Ribose Inc.