leptris 1.9.188.0 → 1.9.193.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: cf80ccf870ddf66abc6a3a4fa99727a9f15d8da89649f337c34d7d24349eb94f
-  data.tar.gz: 95bc3a48e901eda33ba20492d98007da08746dbbef90a391fe518ec37db2942b
+  metadata.gz: d8b1ffa2c04d15482319b91021017e096e8199d6919071ff94c8532af66f0b4d
+  data.tar.gz: bcbbca3c40e50c5d4c42c167f991267f8a670730dfe321c586292d0e201de0f6
 SHA512:
-  metadata.gz: fa413d38c0dc9571c72564d8df2277ddc4c47b0b9a3e38c00824c54b50042042f2594c2acddc94800ba7cd5b455b9b72296b69abe61c0e99711f860de4bd1a71
-  data.tar.gz: '089846a5778f5e3fe3bdf8daa0460f09e81cd4d31250860cf12129de8918caefddc219eb11e445898ca183c94490fed804ed3a89b0bea4fb5c94af6a55eaf8e4'
+  metadata.gz: 7d0ff9f39131c9f66ed2e9f9cc41a266160f917a6edd90c15629e468e88230de2ad68bcaaecfdd7f60ddb62eec4888ab1d2a2800bcb32c77b27cfd01fef49a31
+  data.tar.gz: 04ca510ed71d9d59b58d1b49f9d3875fe134d853612fc7e582889e8821902440de01f096b38dad93a7c9279afa510b2e311790e919ee322b7ba77a07ebd74774

data/CHANGELOG.md

@@ -5,6 +5,33 @@ 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).
 
+## [1.9.192.0] - 2026-09-17
+
+### Changed — libleptris 1.9.188 → 1.9.192 (lockstep)
+
+- **`RelaxNG::Schema#validate_report`** — the whole validation
+  report in one C call (upstream 1.9.190,
+  `leptris_rng_error_report`): `[{ kind:, message:, offender:,
+  line:, column: }]`, empty when valid. `kind` is the #1126
+  failure-class taxonomy ("missing-required-attr",
+  "attr-not-allowed", ...), `offender` the attributed element/
+  attribute name. Spec-pinned. `validate_errors` now renders from
+  the same report (one call + struct reads instead of 4 accessor
+  calls per error); the per-index accessors stay attached for
+  back-compat.
+- **The metanorma schema chain compiles and validates end to end**
+  (1.9.189): the combine-merge heap-use-after-free on include
+  merges, XSD `\i`/`\c` pattern classes, and the whole-value match
+  anchor riding the continuation. Spec-pinned (combine merge,
+  `\i\c*`).
+- **Attributes behind a `<ref>` are consumed** — both the verdict
+  (1.9.191) and the diagnostic walk (1.9.192): basicdoc.rng's
+  `Root-Attributes` no longer produces false
+  "found attribute X, but no attributes allowed here". Spec-pinned.
+- **externalRef resolves in bare-`<element>` schemas** (1.9.190).
+  Spec-pinned.
+- One new public symbol: audit 323/323.
+
 ## [1.9.188.0] - 2026-09-17
 
 ### Changed — libleptris 1.9.186 → 1.9.188 (lockstep)

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 = "1.9.188"
+LIBLEPTRIS_VERSION = "1.9.193"
 # Vendored alongside libleptris for fn:normalize-unicode (TODO
 # .restructure/20): built per platform with a RELOCATABLE @rpath
 # install name, loaded by ffi.rb before libleptris so the
@@ -98,10 +98,59 @@ task :compile do
   # literal on Windows shells (they broke the MSVC build).
   # GCC-family only — MSVC's cl rejects the flag outright (D8021).
   cflags = Gem.win_platform? ? "" : "-Wno-error=incompatible-pointer-types"
-  sh "cmake -B #{build}/build -S #{build} " \
-     "#{CMAKE_FLAGS.join(' ').sub('-DLEPTRIS_ENABLE_UTF8PROC=OFF', '-DLEPTRIS_ENABLE_UTF8PROC=ON')} " \
-     "-DCMAKE_PREFIX_PATH=#{u8_prefix} #{cflags.empty? ? '' : "-DCMAKE_C_FLAGS=#{cflags}"}"
-  sh "cmake --build #{build}/build --config Release -j 4"
+  cmake_base =
+    "#{CMAKE_FLAGS.join(' ').sub('-DLEPTRIS_ENABLE_UTF8PROC=OFF', '-DLEPTRIS_ENABLE_UTF8PROC=ON')} " \
+    "-DCMAKE_PREFIX_PATH=#{u8_prefix} #{cflags.empty? ? '' : "-DCMAKE_C_FLAGS=#{cflags}"}"
+
+  # Two-stage PGO (engine CMake: LEPTRIS_ENABLE_PGO GENERATE/USE;
+  # measured ~20% CPU on the DOM parse path, v1.9.188 worktree A/B,
+  # every PGO round beating every baseline round). Train via the CLI
+  # over the tarball's XML corpus. Windows keeps the plain build —
+  # MSVC PGO is inherently LTCG and these gems link the DLL as-is.
+  # LEPTRIS_PGO=0 escapes to the single-stage build.
+  pgo = ENV["LEPTRIS_PGO"] != "0" && !Gem.win_platform?
+  if pgo
+    # _GNU_SOURCE for the TRAIN build only: the CLI (the trainer, not
+    # built in normal gems) uses fileno, which musl hides without a
+    # POSIX feature macro (glibc is lenient). Upstream fix pending in
+    # the C repo; this stays on the throwaway instrumented build.
+    # _GNU_SOURCE is MUSL-ONLY: the CLI trainer's fileno needs a POSIX
+    # feature macro under musl. On glibc the macro reshuffles the
+    # header graph and breaks arena.c's accidental (missing-stdint.h)
+    # uintptr_t. Single-quote: the value carries a space (an unquoted
+    # split hands cmake a bare -D_GNU_SOURCE); non-Windows branch, so
+    # sh quoting is safe.
+    gnu = RUBY_PLATFORM =~ /musl/ ? " -D_GNU_SOURCE" : ""
+    pgo_cflags = "#{cflags}#{gnu}".strip
+    pgo_base =
+      if pgo_cflags == cflags
+        cmake_base
+      else
+        cmake_base.sub("-DCMAKE_C_FLAGS=#{cflags}",
+                       "-DCMAKE_C_FLAGS='#{pgo_cflags}'")
+      end
+    sh "cmake -B #{build}/build -S #{build} #{pgo_base} " \
+       "-DLEPTRIS_BUILD_CLI=ON -DLEPTRIS_ENABLE_PGO=GENERATE"
+    sh "cmake --build #{build}/build --config Release -j 4"
+    cli = File.join(build, "build", "cli", "leptris")
+    pgo_dir = File.join(build, "build", "pgo-data")
+    corpus = Dir[File.join(build, "benchmarks", "data", "*.xml")]
+    corpus.each do |xml|
+      # a corpus file can be deliberately malformed; training tolerates it
+      system({ "LLVM_PROFILE_FILE" => File.join(pgo_dir, "%p.profraw") },
+             cli, "parse", xml, out: File::NULL, err: File::NULL)
+    end
+    if RUBY_PLATFORM =~ /darwin/
+      sh "xcrun llvm-profdata merge -output=#{File.join(pgo_dir, 'default.profdata')} " \
+         "#{File.join(pgo_dir, '*.profraw')}"
+    end
+    sh "cmake -B #{build}/build -S #{build} #{cmake_base} " \
+       "-DLEPTRIS_BUILD_CLI=OFF -DLEPTRIS_ENABLE_PGO=USE"
+    sh "cmake --build #{build}/build --config Release -j 4"
+  else
+    sh "cmake -B #{build}/build -S #{build} #{cmake_base}"
+    sh "cmake --build #{build}/build --config Release -j 4"
+  end
   # Windows names the shared library leptris.dll (no "lib" prefix);
   # vendoring under the uniform libleptris.* name keeps the FFI
   # search order simple. libleptris >= 1.3.0 renames the DLL to

data/lib/leptris/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Leptris
-  VERSION = "1.9.188.0"
+  VERSION = "1.9.193.0"
 end

data/lib/leptris/xml/ffi.rb

@@ -1027,6 +1027,21 @@ attach_function :leptris_parse_string,
       attach_function :leptris_rng_error_column,
         [:leptris_relaxng, :size_t], :int
 
+      # Whole-report face (libleptris 1.9.190): one call returns
+      # every record — kind (the #1126 failure-class taxonomy) and
+      # offender were unreachable through the per-index accessors.
+      # Record array is handle-owned; strings live until the next
+      # validate call or leptris_rng_free.
+      class RngErrorRecord < ::FFI::Struct
+        layout :kind, :string,
+               :message, :string,
+               :offender, :string,
+               :line, :uint,
+               :column, :uint
+      end
+      attach_function :leptris_rng_error_report,
+        [:leptris_relaxng, :pointer], :size_t
+
       attach_function :leptris_xslt_parse,
         [:string, :size_t], :leptris_xslt
       attach_function :leptris_xslt_parse_file,

data/lib/leptris/xml/relaxng.rb

@@ -23,6 +23,7 @@ module Leptris::XML::RelaxNG
   #     rng.valid?(doc)      # => true/false
   #     rng.validate(doc)    # => [] or ["1:0: error: element ..."]
   #     rng.validate_errors(doc) # => [] or [{line:, column:, message:}]
+  #     rng.validate_report(doc) # => [] or [{kind:, message:, offender:, ...}]
   #
   class Schema
     # GC-managed compiled handle.
@@ -86,23 +87,40 @@ module Leptris::XML::RelaxNG
     # callers need for log formatting, instead of re-parsing the
     # Jing-form strings from #validate. Empty when valid. Message,
     # line, and column carry Jing's exact attribution (libleptris
-    # >= 1.9.179, upstream #878). This is the single enumeration of
-    # the error-accumulation surface; #validate renders these rows.
+    # >= 1.9.179, upstream #878). Rendered from #validate_report —
+    # the single enumeration of the error surface.
     def validate_errors(document)
+      validate_report(document).map do |r|
+        { line: r[:line], column: r[:column], message: r[:message] }
+      end
+    end
+
+    # The whole validation report, one C call (libleptris >= 1.9.190):
+    # [{ kind:, message:, offender:, line:, column: }] — empty when
+    # valid. `kind` is the #1126 failure-class taxonomy
+    # ("missing-required-attr", "attr-not-allowed", ...), `offender`
+    # the attributed element/attribute name (nil when unknown).
+    # Strings are read into Ruby copies before returning.
+    def validate_report(document)
       ok = Leptris::XML::FFI.leptris_rng_validate(@handle,
                                                   document.c_ptr)
       return [] if ok != 0
-      count = Leptris::XML::FFI.leptris_rng_error_count(@handle)
+      out = ::FFI::MemoryPointer.new(:pointer)
+      count = Leptris::XML::FFI.leptris_rng_error_report(@handle, out)
       return [] if count.zero?
+      base = out.read_pointer
+      return [] if base.null?
+      record = Leptris::XML::FFI::RngErrorRecord
       Array.new(count) do |i|
-        msg = Leptris::XML::FFI.leptris_rng_error_message(@handle, i)
-        next nil if msg.nil?
+        r = record.new(base + i * record.size)
         {
-          line: Leptris::XML::FFI.leptris_rng_error_line(@handle, i),
-          column: Leptris::XML::FFI.leptris_rng_error_column(@handle, i),
-          message: msg,
+          kind: r[:kind],
+          message: r[:message],
+          offender: r[:offender],
+          line: r[:line],
+          column: r[:column],
         }
-      end.compact
+      end
     end
   end
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: leptris
 version: !ruby/object:Gem::Version
-  version: 1.9.188.0
+  version: 1.9.193.0
 platform: ruby
 authors:
 - Ribose Inc.