leptris 1.9.42 → 1.9.44

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 5de687a1b7597e0bdf6ff80f85566435d9927e136a86ec758a554928d85910d2
-  data.tar.gz: 9171a19d4620e4d482c573e1f21e76a8ee253b175a8df34569e9711f1fcc42e4
+  metadata.gz: 9895934a1c086857555795905c15ab4f74ef053804e489674192bc7f4cc70bf3
+  data.tar.gz: b1bf31ac8c0ef5f718d78d4ae0654218cd3c212adf33e75628326949644032d1
 SHA512:
-  metadata.gz: b2dadeb0a139c9a6396dd82defd080dc014e056a8959321a7f7dff13851503d8d53b6a7b1dc4f987a4bb131ce9054e3c0073d7234d6a8ff0e73caae2fc71cf7e
-  data.tar.gz: 1648229492701f2805e405475b7abf4f59108844d79da56118f47e15fdb048ebf03f5b7a1d4c3f48c566e2c13b1881b50b7a09df9512fcb2cd928ae045260f3a
+  metadata.gz: 536c0b5a7ddb425395b077f5a6a0b40ebc893c6037ba71277a1b6eaa70b9769e977b03b9ed0f81ccf6d9a70e11c4f4095b8ab3e473268f2cb38306ea897d28e4
+  data.tar.gz: d34c001bf6840c12d3e44a0253166938523f57c0e935a6dfa4560cea575b02af043d25caf0a491a77810726b168c64e556c63b825355eb3ff23ffbbf010848c5

data/CHANGELOG.md

@@ -5,6 +5,61 @@ All notable changes to Leptris will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.9.44] - 2026-08-31
+
+### Added
+
+- **libleptris 1.9.19–1.9.22 lockstep** (audit 255/255, three new
+  symbols): the #653 hold is lifted — the reported parse
+  "regression" was an ill-formed fixture (stray `</y>`) plus a
+  stale-build bisect artifact; forced-clean rebuilds show 1.9.18
+  and 1.9.21+ identical on every well-formed shape (verified and
+  withdrawn on the issue).
+- **`Node#visit`** (upstream #645a, `leptris_node_visit`): one C
+  call over the subtree — elements yield `(node, true/false,
+  depth)` enter/leave pairs, other kinds once, depth from the
+  receiver; the document node walks the document chain. No
+  NodeSet, pointer array, or children memo per level — the leanest
+  full-subtree iteration the binding offers (#645a's walk lever;
+  consumers can now beat the children-recursion allocation floor).
+- **Indent unit with Nokogiri semantics (leptris-ruby#109,
+  upstream #633)**: `to_xml(indent: 2, indent_text: "\t")` — the
+  unit string replaces the default spaces, repeated `indent`
+  times per depth level, standard layout otherwise
+  (byte-identical to Nokogiri's output modulo their trailing
+  newline, which moxml normalizes). `indent_text: true` keeps the
+  display form; `false` the default. Wired through the sized
+  ext-serialize entry (PR #107's substance — the audit requires it
+  attached with the lockstep, so it rides here; the engine emits
+  one unit copy per level and the binding multiplies unit x indent
+  to reach Nokogiri's repeat count).
+
+## [1.9.43] - 2026-08-31
+
+### Fixed
+
+- **Pull prefix events surface (leptris-ruby#648's residual)**: the
+  engine's pull stream has always carried start/end prefix events
+  (codes 8/9; the default namespace's prefix is legitimately `""`)
+  — the binding's type map lacked the kinds, silently delivering
+  nil-typed Events that the 1.9.42 staging guard then misread as
+  corruption on healthy engines. `:start_prefix`/`:end_prefix` now
+  map correctly on both cursor and batch paths, and the guard's
+  name check only invalidates kinds that must carry names
+  (start/end element, PI target). Engine 1.9.21's staging fix
+  (upstream #648) verified clean through the batch path once the
+  kinds were mapped.
+
+### Meta
+
+- The 1.9.19–1.9.21 lockstep is HELD: engine 1.9.19+ carries a DOM
+  parse regression — a self-closing element followed by text inside
+  a depth-2 parent fails to parse (`<div><p><br/>hello</p></div>`)
+  — filed as leptris/leptris#653. `Node#visit` (the #645a
+  wrap-free walk), the sized ext-serialize entry (PR #107's
+  substance), and this pull fix's engine-1.9.21 verification are
+  parked on a branch awaiting the fix.
+
 ## [1.9.42] - 2026-08-30
 
 ### Fixed

data/Rakefile

@@ -8,7 +8,7 @@ RSpec::Core::RakeTask.new(:spec)
 # Pin for `rake compile` and the platform-gem builds. Keep in lockstep
 # with .github/workflows/build.yml (which calls `rake compile`) and the
 # CHANGELOG when libleptris releases.
-LIBLEPTRIS_VERSION = "1.9.18"
+LIBLEPTRIS_VERSION = "1.9.22"
 
 CMAKE_FLAGS = %w[
   -DCMAKE_BUILD_TYPE=Release

data/lib/leptris/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Leptris
-  VERSION = "1.9.42"
+  VERSION = "1.9.44"
 end

data/lib/leptris/xml/document.rb

@@ -266,13 +266,21 @@ class Leptris::XML::Document
   end
   alias_method :internal_subset, :doctype
 
-  # indent_text: true selects the display form (libleptris 1.9.9,
-  # #129): text and mixed content indent too — the output is
-  # display-oriented and not round-trip-guaranteed.
+  # indent_text selects the ext-serializer knob (libleptris 1.9.22):
+  # a STRING is the indent unit with Nokogiri's semantics — the unit
+  # replaces the default spaces, one copy per depth level, standard
+  # layout (requires indent > 0); true selects the display form
+  # (1.9.9 #129 — text and mixed content indent too; output is
+  # display-oriented and not round-trip-guaranteed).
   def to_xml(indent: 0, no_decl: false, encoding: nil, indent_text: false)
     raise Leptris::XML::UseAfterFreeError if @freed.state == :freed
     return "" if @c_ptr.nil?
-    if indent_text
+    case indent_text
+    when String
+      return Leptris::XML::Serialization.to_xml_indent_unit(
+        @c_ptr, indent_text, indent: indent, no_decl: no_decl,
+        encoding: encoding)
+    when true
       return Leptris::XML::Serialization.to_xml_display(
         @c_ptr, indent: indent, no_decl: no_decl, encoding: encoding)
     end

data/lib/leptris/xml/ffi.rb

@@ -72,10 +72,13 @@ module Leptris
           :encoding, :pointer
       end
 
-      # LeptrisSerializeExtOptions (1.9.9, #129): display-form
-      # serialization knobs beyond the ABI-frozen options struct.
+      # LeptrisSerializeExtOptions (1.9.9 #129 display form; 1.9.22
+      # #633/#109 indent unit): knobs beyond the ABI-frozen options
+      # struct. MAY grow — always pass its size through the sized
+      # entry.
       class SerializeExtStruct < ::FFI::Struct
-        layout :indent_text, :int
+        layout :indent_text, :int,
+               :indent_unit, :pointer
       end
 
       # Mirrors LeptrisParseOptions (libleptris >= 1.9.0 carries the
@@ -154,6 +157,12 @@ module Leptris
       # 1.9.7, upstream #580; the libxml2 model).
       attach_function :leptris_document_node,
         [:leptris_document], :leptris_node_ref
+      # Wrap-free subtree visitation (1.9.20, upstream #645a):
+      # enter/leave pairs for elements, depth from the walk's root,
+      # one C call — no NodeSet/Array churn per level. The document
+      # node walks the document child chain.
+      attach_function :leptris_node_visit,
+        [:leptris_node_ref, :pointer, :pointer], :void
       # Document-level PIs (v1.6.0): not tree nodes — enumerate via
       # these accessors only.
       attach_function :leptris_document_pi_count,
@@ -646,6 +655,11 @@ module Leptris
       # read_owned_string (the seam frees it).
       attach_function :leptris_document_serialize_ext,
         [:leptris_document, :pointer, :pointer], :pointer
+      # Sized variant (1.9.19, upstream #644 — leptris-ruby PR #107):
+      # passing our ext-struct size keeps the engine from reading
+      # past this binding's allocation (the MSVC segfault).
+      attach_function :leptris_document_serialize_ext_sized,
+        [:leptris_document, :pointer, :pointer, :size_t], :pointer
       attach_function :leptris_document_serialize_into,
         [:leptris_document, :pointer, :size_t, :pointer, :pointer], :size_t
       attach_function :leptris_document_get_dtd,
@@ -766,6 +780,10 @@ module Leptris
       PULL_PI            = 5
       PULL_END_DOCUMENT  = 6
       PULL_ERROR         = 7
+      # Prefix events (types.h LeptrisPullEventType 8/9): the
+      # default namespace's prefix is legitimately "".
+      PULL_START_PREFIX  = 8
+      PULL_END_PREFIX    = 9
 
       LEPTRIS_PARSE_DEFAULT = 0
       LEPTRIS_PARSE_DROP_WS_TEXT = 1

data/lib/leptris/xml/node.rb

@@ -273,6 +273,31 @@ class Leptris::XML::Node
   end
   alias_method :remove, :unlink
 
+  # Visits the subtree with ONE C call (leptris_node_visit,
+  # libleptris 1.9.20 — upstream #645a): elements yield twice —
+  # (node, true, depth) before their children, (node, false, depth)
+  # after the subtree completes — every other kind once with
+  # entering=true; depth counts element levels from the receiver.
+  # No NodeSet, pointer array, or children memo per level: the
+  # leanest full-subtree iteration the binding offers (a document
+  # receiver walks the document child chain). The walk is read-only
+  # — mutate only between visits.
+  #
+  #     root.visit { |node, entering, depth| ... }
+  def visit(&block)
+    return enum_for(:visit) unless block
+    ensure_alive!
+    document = @document
+    visitor = ::FFI::Function.new(
+      :void, [:pointer, :pointer, :int, :int], blocking: true) do |_, node_ptr, entering, depth|
+      block.call(
+        Leptris::XML::Node.wrap(node_ptr, document),
+        entering == 1, depth)
+    end
+    Leptris::XML::FFI.leptris_node_visit(@c_ptr, visitor, nil)
+    self
+  end
+
   # Walks the subtree in post-order DFS (matches Nokogiri's
   # semantics): the receiver, its descendants, nothing else.
   #

data/lib/leptris/xml/pull.rb

@@ -28,6 +28,13 @@ module Leptris::XML::Pull
     Leptris::XML::FFI::PULL_PI => :pi,
     Leptris::XML::FFI::PULL_END_DOCUMENT => :end_document,
     Leptris::XML::FFI::PULL_ERROR => :error,
+    # Prefix events ride the stream ("" prefix = the default
+    # namespace): the engine has always delivered them; the map
+    # lacked the kinds, silently yielding nil-typed Events — which
+    # the staging guard then misread as corruption after the real
+    # engine fix landed (leptris/leptris#648's residual).
+    Leptris::XML::FFI::PULL_START_PREFIX => :start_prefix,
+    Leptris::XML::FFI::PULL_END_PREFIX => :end_prefix,
   }.freeze
 
   class Parser
@@ -87,7 +94,7 @@ module Leptris::XML::Pull
           type = TYPES[buffer.get_int(base + type_off)]
           name = name_ptr.null? ? nil :
             name_ptr.read_string.force_encoding(Encoding::UTF_8)
-          # Staging-corruption guard (leptris/leptris#646): the C
+          # Staging-corruption guard (leptris/leptris#648): the C
           # batch's staging arena misplaces record strings once the
           # staged content crosses its block boundary (~185 bytes of
           # attribute value under nested attr-carrying ancestors) —
@@ -97,10 +104,12 @@ module Leptris::XML::Pull
           # instead of delivering garbage, and point at the cursor
           # path, which reads the same documents correctly.
           if type.nil? ||
-             (name && (name.empty? || !name.valid_encoding? ||
-                       name.match?(/[\x00-\x08\x0e-\x1f]/)))
+             (name && (!name.valid_encoding? ||
+                       name.match?(/[\x00-\x08\x0e-\x1f]/))) ||
+             (%i[start_element end_element pi].include?(type) &&
+              (name.nil? || name.empty?))
             raise Leptris::XML::Error,
-              "pull batch staging corruption (leptris/leptris#646) — " \
+              "pull batch staging corruption (leptris/leptris#648) — " \
               "use Parser#each (cursor) for this document"
           end
           text = text_ptr.null? ? nil :

data/lib/leptris/xml/serialization.rb

@@ -47,8 +47,28 @@ module Leptris::XML::Serialization
   def self.to_xml_display(c_ptr, indent: 0, no_decl: false, encoding: nil)
     opts, encoding_anchor = build_options(
       indent: indent, no_decl: no_decl, encoding: encoding)
-    str_ptr = Leptris::XML::FFI.leptris_document_serialize_ext(
-      c_ptr, opts.pointer, INDENT_TEXT_EXT.pointer)
+    str_ptr = Leptris::XML::FFI.leptris_document_serialize_ext_sized(
+      c_ptr, opts.pointer, INDENT_TEXT_EXT.pointer, INDENT_TEXT_EXT.size)
+    Leptris::XML::FFI.read_owned_string(str_ptr)
+  end
+
+  # Indent-unit serialization (libleptris 1.9.22, #633 —
+  # leptris-ruby#109): Nokogiri's indent_text semantics — the unit
+  # string replaces the default spaces, repeated `indent` times per
+  # depth level, standard layout otherwise. The engine emits ONE
+  # copy of the unit string per level (options->indent is ignored
+  # when a unit is set), so the binding multiplies unit x indent to
+  # reach Nokogiri's repeat count.
+  def self.to_xml_indent_unit(c_ptr, unit, indent: 0, no_decl: false,
+                              encoding: nil)
+    opts, encoding_anchor = build_options(
+      indent: indent, no_decl: no_decl, encoding: encoding)
+    ext = Leptris::XML::FFI::SerializeExtStruct.new
+    unit_anchor = ::FFI::MemoryPointer.from_string(
+      indent.positive? ? unit.to_s * indent : unit.to_s)
+    ext[:indent_unit] = unit_anchor
+    str_ptr = Leptris::XML::FFI.leptris_document_serialize_ext_sized(
+      c_ptr, opts.pointer, ext.pointer, ext.size)
     Leptris::XML::FFI.read_owned_string(str_ptr)
   end
 

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: leptris
 version: !ruby/object:Gem::Version
-  version: 1.9.42
+  version: 1.9.44
 platform: ruby
 authors:
 - Ribose Inc.