leptris 1.9.374.0-aarch64-linux → 1.9.375.0-aarch64-linux

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f1cfdde9b79dc733fd841f7c812984e03a3b26da42ec56232dc0847d840fe7f6
-  data.tar.gz: a620bff9d9f386b970c6f7523119a5a7635e3d6d61b2d3d9170a8a1ac17ab6d4
+  metadata.gz: ced7790ba0349b52901cfe048ccaaee9c173b4b9b1de734c0d3fabd33a16cd8f
+  data.tar.gz: 67b2f8ad00b72be5df814d61a54beafd7d3aca4e1e11990eb82c6e4d2111d760
 SHA512:
-  metadata.gz: 85e7d38d6393fd494646d20630b3bd31b8e35536c47b22b83c12277ecc6a431b82255e0cf191aeb3c87f5a5dd824fa064113d88a4e9c37989b9ee702f26676e5
-  data.tar.gz: 1913d5728de2ae72375de606fb48ca5b09b9964775468d5cf2f6c273d8079f9b359c71e34fc99eb377307aa664abebd519716b9285a2230543e1bce1d5bf65eb
+  metadata.gz: 897ed4da545dd52d98e29024bd707e19f8de4f9339e2b3febfe315332595fa05e7fb3868960adcd7c3d867b359e8eb6e1ed5e9b85910e856601b33ef6f400f88
+  data.tar.gz: 4234b33c3e5b9a2c5a1bfb219ce73a8fcda24afe4a1415d89eaa930df5ac16776231d47db8300a39ebd1a580d398cbe941a02d4e35f4b0b8718981ed1644c9f7

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 = "1.9.374"
+LIBLEPTRIS_VERSION = "1.9.375"
 # 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

data/lib/leptris/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Leptris
-  VERSION = "1.9.374.0"
+  VERSION = "1.9.375.0"
 end

data/vendor-src/README.md

@@ -1,7 +1,7 @@
 Sources vendored in this gem (recompile rights; the ruby
 variant builds them at install via extconf.rb):
 
-- libleptris v1.9.374
+- libleptris v1.9.375
 - utf8proc v2.11.0
 
 Rebuild by hand:

data/vendor-src/libleptris/CHANGELOG.md

@@ -1,5 +1,22 @@
 ## [Unreleased]
 
+## [1.9.375] - 2026-10-20
+
+### Added
+
+- **html: the #1771 verdict-parity emission slices (PRs #1788,
+  #1789, #1790, #1791) — parity 1394 to 1423 of 1555.** The EOF
+  closing-tag verdict now implements html5lib's InBody processEOF
+  allowed-set exactly (zero false positives — the lint's first
+  requirement); the in-select drop sites report
+  unexpected-start/end-tag-in-select and unexpected-select-in-
+  select; the foster branch reports foster-parenting-character/
+  -start-tag by node kind; unknown-doctype follows html5lib's
+  BeforeHtml processDoctype (name/public/system shape). Trees
+  byte-identical throughout (1555/1555); full suite green at
+  every step.
+
+
 ## [1.9.374] - 2026-10-19
 
 ### Fixed

data/vendor-src/libleptris/CMakeLists.txt

@@ -3,7 +3,7 @@
 cmake_minimum_required(VERSION 3.20)
 
 project(leptris
-    VERSION 1.9.374
+    VERSION 1.9.375
     DESCRIPTION "Fast XML parser and XPath evaluator in pure C"
     LANGUAGES C CXX
 )

data/vendor-src/libleptris/TODO.SOTA.md

@@ -0,0 +1,154 @@
+# TODO.SOTA — the vision of done
+
+The audit underneath this board: an XML stack is an **epistemic
+instrument** — it decides what a document means. Its purpose is
+trustworthy interpretation, made affordable everywhere. SOTA is
+therefore not a feeling but four provable claims, each with a
+measurement command whose output is the verdict:
+
+1. **Conformance** — we interpret every document the way the
+   specifications (and the reference suites) say. *Proof: zero reds
+   on every adopted suite.*
+2. **Performance** — correctness is free; we are faster than every
+   incumbent on every published shape. *Proof: ≥1.00x on every row,
+   interleaved medians.*
+3. **Safety** — trust is auditable: memory-safe under sanitizers,
+   fuzz-clean, reproducibly built. *Proof: clean lanes + per-release
+   build manifests.*
+4. **Universality** — the same semantics in every language; a
+   document means the same thing through every runtime. *Proof: a
+   cross-runtime differential corpus, byte-identical.*
+
+Everything below is the task board for those four proofs. Every row
+ends with a dated measurement; the scoreboard lives in
+docs/MISSION-PLAN.md and is updated at every release.
+
+---
+
+## Gate A — Conformance: zero reds everywhere
+
+### A1 WHATWG verdict parity: 1394/1555 → 1555/1555
+The remaining 161 verdict cases + ~270 code-set divergences
+(html5lib-verdict-redlist.txt). Ground every slice in html5lib's
+actual source (~/src/external/html5lib):
+- [ ] escaped-state EOF codes (`unexpected-eof-in-text-mode`,
+      `eof-in-script-in-script`) — the script-data state machine's
+      EOF classification
+- [ ] the after-body EOF phase rule (tests1:34 class; the exact
+      phaseAtBody.eof semantics)
+- [ ] in-body `unexpected-start-tag` family (body/frameset/form
+      starts; the startTagBody/Forms/Frameset error sites)
+- [ ] `unexpected-end-tag` family (p-in-button-scope synthesis,
+      form pointer, li/dd/dt scope, endTagOther walk)
+- [ ] foster-parenting sites (start-tag/character/character-in-table)
+- [ ] adoption-agency codes (`adoption-agency-1.3` et al.)
+- [ ] in-table voodoo (`*-implies-table-voodoo`, cells, eof-in-table)
+- [ ] unknown-doctype / unexpected-char-in-doctype (doctype
+      tokenization §13.2.5.28+)
+- [ ] non-html-root, select-mode codes, heading/edge shapes
+
+### A2 QT3 / XQuery 3.1: zero unexplained pins
+- [ ] re-pin fresh (the ~41–99 slice figures are dated)
+- [ ] every remaining pin gets a disposition file or a fix
+- [ ] ctor node-yield family: owned_docs transfer completion
+
+### A3 XSD conformance lane: suite-complete
+- [ ] the lane's own queue (769→795 → its ceiling) to zero
+
+### A4 DTD phase 8 (TODO 91)
+- [ ] ENTITY/ENTITIES attribute resolution
+- [ ] parameter entities, choice-model backtracking
+
+### A5 XInclude completeness (TODO 92)
+- [ ] `parse="xml"` (ownership-transfer splice exists — wire it)
+- [ ] xpointer
+
+### A6 The universality proof (differential corpus)
+- [ ] one curated corpus (the html5lib + QT3-winner + metanorma
+      sample set) run through ruby/py/ts/rs bindings
+- [ ] byte-identical outputs asserted in CI (the "one engine, same
+      numbers everywhere" proof)
+
+### A7 Maintain (already green; keep gated)
+- [ ] XML 1.0, XPath 438/438, libxslt 205/205, XsltFull,
+      C14N, RelaxNG, SAX records lane
+
+## Gate B — Performance: every published row ≥1.00x
+
+### B1 build row: ~2.9x → ≤1.0x (the #242 campaign)
+Harness: wt/460-harness/build_bench.rb. Decomposition on record.
+- [ ] per-element object budget: one Ruby object per created node
+      (drop the shell entirely post-create-priming; the mint cache
+      insert policy; the context registration)
+- [ ] the root `to_binding` bridge (set_root path)
+- [ ] `[]=` / `text=` write faces (nn_set_attribute pattern applied
+      to text writes; one C frame per write)
+
+### B2 walk row: 0.88x → much faster
+Per-face reads already 3–5x ahead; the harness floor is shared
+scaffolding. Memo law holds (dropping @children memo = 2x GC).
+- [ ] hot-path audit of NodeSet#each / block dispatch
+- [ ] publish the per-face table (the honest much-faster story)
+
+### B3 whole-pipeline rows (new, published)
+- [ ] parse→query→serialize pipeline row vs nokogiri/libxml2
+- [ ] build→transform (XSLT) pipeline row
+- [ ] both to serialbench as the public scoreboard
+
+### B4 durability
+- [ ] CI perf regression gates on the chokepoints (parse, walk,
+      build, serialize medians, fail on >5% regression)
+
+### B5 remaining red rows
+- [ ] pugixml lane-18 append/attr-parse (2.0x/5.9x; hash prefilter)
+- [ ] zero-copy text lever (#1218) — adopt if it wins any row
+
+## Gate C — Safety: trust is auditable
+
+- [ ] C1 structure-aware fuzzing: grammar-aware corpora for
+      XML/XPath/XSLT/HTML (grow beyond the three libFuzzer harnesses;
+      corpus commit + crash-triage SLA)
+- [ ] C2 sanitizer lanes over the risk families: records-drain,
+      borrowed-slice + pool-recycle (the #506 class), iterparse
+      keepalive — ASAN/UBSAN now, add TSAN (batch/worker surfaces)
+- [ ] C3 reproducible builds everywhere: darwin toolchain pin done
+      (ruby#512) — extend to linux/windows legs; publish a per-
+      release build manifest (engine + binding hashes per platform)
+- [ ] C4 security.md: advisory channel, CVE-response process,
+      dependency-free attestation
+- [ ] C5 the memory story documented: pool arena model, leak gates,
+      the borrowed-slice lifetime contract
+
+## Gate D — Universality: same semantics everywhere
+
+- [ ] D1 leptris-descriptor published (crates.io; feat/294 branch
+      lands), rides lockstep
+- [ ] D2 system packaging: homebrew, conan, deb (apt repo)
+- [ ] D3 cross-runtime differential CI (= A6, wired)
+- [ ] D4 libxml2→libleptris migration guide (docs/guide/) + the
+      nokogiri-compat surface documented
+
+## Gate E — 1.0 (the seal)
+
+- [ ] E1 Gate A + B report zero reds (dated scoreboard)
+- [ ] E2 semver freeze, ABI stability policy, platform matrix doc
+- [ ] E3 v1.0.0 through the release workflow — never tagged by hand
+
+---
+
+## Invariants (the laws the board executes inside)
+
+- Falsifiability: red-first specs; both-directions measurement.
+- Lockstep + TAG-RACE; the release workflow owns versions and tags.
+- One branch/PR per item; worktrees at wt/; never commit on main;
+  never `git add -A`; verify the branch before every commit.
+- Aliases before shims: unshadowed C faces upstream; consumers
+  face-probe and plain-dispatch.
+- Suites over opinions; the corpus is the reference.
+
+## Definition of done
+
+Every checkbox above carries a dated measurement command and its
+recorded result; docs/MISSION-PLAN.md's scoreboard shows Gate A and
+Gate B at zero reds; v1.0.0 is tagged. That day, the libxml2 era is
+over wherever this engine ships.

data/vendor-src/libleptris/src/leptris/html/html_parse.c

@@ -4136,6 +4136,16 @@ static void h_append(HBuilder* b, LeptrisNodeRef n) {
         /* #659 foster (WHATWG only): text/elements in table context
          * go before the nearest open table in ITS parent. */
         if (b->whatwg_foster && tbl_ctx && h_fosterable(b, n)) {
+            /* #1771: the corpus verdicts name the foster by node
+             * kind (foster-parenting-character / -start-tag). The
+             * position rides the line memo — h_append has no token
+             * offset, and the verdict gate compares codes only. */
+            h_diag(b,
+                   leptris_node_get_type(n) == LEPTRIS_NODE_TYPE_TEXT
+                       ? "foster-parenting-character"
+                       : "foster-parenting-start-tag",
+                   b->err_line_memo_off,
+                   "node fostered out of the table context");
             int ti = (int)b->depth - 1;
             while (ti >= 0 && !h_ieq_raw(
                        leptris_element_name(b->open[ti]), "table"))
@@ -5903,6 +5913,21 @@ static LeptrisDocument html_parse_shared(
                      * doctype - with an EMPTY name (13.2.5 "missing-
                      * doctype-name", doctype01.dat:4/5). WHATWG
                      * mode only; html4 keeps the drop. */
+                    /* #1771: html5lib's BeforeHtml processDoctype —
+                     * name != html, a public id, or a system id other
+                     * than about:legacy-compat is unknown-doctype. */
+                    if (b.whatwg) {
+                        size_t ln = nlen;
+                        int name_html = (ln == 4);
+                        for (int i = 0; name_html && i < 4; i++)
+                            if (h_lower(nstart[i]) != "html"[i])
+                                name_html = 0;
+                        if (!name_html || pub[0] ||
+                            (sys[0] && strcmp(sys, "about:legacy-compat") != 0))
+                            h_diag(&b, "unknown-doctype",
+                                   (size_t)(nstart - b.input_base),
+                                   "unknown DOCTYPE shape");
+                    }
                     if (nlen || b.whatwg) {
                         /* WHATWG lowercases the doctype name;
                          * libxml2 (the html4/parity mode) preserves
@@ -6357,6 +6382,13 @@ static LeptrisDocument html_parse_shared(
                                    strcmp(lname, "select") != 0 &&
 
                                    strcmp(lname, "template") != 0) {
+                            /* #1771: html5lib's InSelect endTagOther
+                             * reports the drop (unexpected-end-tag-
+                             * in-select). */
+                            h_diag(&b, "unexpected-end-tag-in-select",
+                                   (size_t)(p - b.input_base),
+                                   "</%s> ignored inside <select>",
+                                   lname);
                             p = q;
                             text = p;
                             continue;
@@ -7215,7 +7247,11 @@ static LeptrisDocument html_parse_shared(
         if (b.whatwg && h_in_select(&b) &&
             strcmp(name, "select") == 0) {
             /* 13.2.6.4.11: a <select> start tag inside a select
-             * ACTS AS its end tag — no nesting (tests1:30). */
+             * ACTS AS its end tag — no nesting (tests1:30).
+             * #1771: html5lib reports unexpected-select-in-select. */
+            h_diag(&b, "unexpected-select-in-select",
+                   (size_t)(p - b.input_base),
+                   "<select> inside <select> acts as an end tag");
             for (size_t d2 = b.depth; d2 > 0; d2--)
                 if (strcmp(leptris_element_name(b.open[d2 - 1]),
                            "select") == 0) {
@@ -7274,6 +7310,11 @@ static LeptrisDocument html_parse_shared(
                 b.depth = (size_t)sel_idx;   /* pop the select;
                                                * reprocess below */
             } else {
+                /* #1771: html5lib's InSelect startTagOther reports
+                 * the drop (unexpected-start-tag-in-select). */
+                h_diag(&b, "unexpected-start-tag-in-select",
+                       (size_t)(p - b.input_base),
+                       "<%s> ignored inside <select>", name);
                 while (q < end && *q != '>') q++;
                 p = (q < end) ? q + 1 : end;
                 text = p;
@@ -8323,29 +8364,43 @@ done:
     /* Nothing appended (empty input, stray end tags only, doctype
      * only) is NOT an error in lenient HTML mode: the wrapper
      * synthesis below yields the empty Nokogiri shape. */
-    /* #1771: elements still open at EOF are
-     * expected-closing-tag-but-got-eof (html5lib's verdict; the
-     * implied html shell does not count). Judged BEFORE the
-     * wrapper synthesis, which tears the open stack down. Our
-     * stack carries no implied html/body, so the html5lib
-     * top-of-stack rule maps to: any real element open, gated on
-     * </body>/</html> never having been seen — after-body EOF is
-     * clean whatever remains (tests1:34). A raw-text container
-     * got its own variant at the raw branch already (that branch
-     * pops the element, so the top here is whatever wraps it). */
+    /* #1771: html5lib's InBody processEOF (html5parser.py:958):
+     * ONE expected-closing-tag-but-got-eof iff ANY open element is
+     * outside the allowed set {dd,dt,li,p,tbody,td,tfoot,th,thead,
+     * tr,body,html} — li/p/table scaffolding does not count
+     * (tests1:34), address/applet/div do (blocks.dat:1,
+     * tests20:96-class). AfterBody's processEOF is silent
+     * (html5parser.py:2533) and its non-wsCharacters rule flips the
+     * phase back to in-body — the builder's after_body/after_body_done
+     * pair is exactly that state. A raw-text container got its own
+     * variant at the raw branch already (that branch pops the
+     * element). Judged BEFORE the wrapper synthesis tears the open
+     * stack down. */
     if (b.whatwg && b.depth > 0) {
-        uint8_t top = b.open_id[b.depth - 1];
-        if (top != h_id_html && top != h_id_head &&
-            top != h_id_body)
-            h_diag(&b,
-                   (top == h_id_script || top == h_id_style ||
-                    top == h_id_listing || top == h_id_plaintext ||
-                    top == h_id_textarea || top == h_id_title ||
-                    top == h_id_xmp || top == h_id_iframe ||
-                    top == h_id_noembed || top == h_id_noframes)
-                       ? "expected-named-closing-tag-but-got-eof"
-                       : "expected-closing-tag-but-got-eof",
-                   b.input_len, "elements left open at end of input");
+        static uint8_t k_allowed[256];
+        static int allowed_init;
+        if (!allowed_init) {
+            /* id tables are lazily initialized once; these ids are
+             * stable for the process lifetime */
+            uint8_t ids[] = {h_id_p, h_id_li, h_id_dd, h_id_dt,
+                             h_id_tbody, h_id_td, h_id_tfoot,
+                             h_id_th, h_id_thead, h_id_tr,
+                             h_id_body, h_id_html};
+            for (size_t k = 0; k < sizeof(ids) / sizeof(ids[0]); k++)
+                k_allowed[ids[k]] = 1;
+            allowed_init = 1;
+        }
+        for (size_t k = 0; k < b.depth; k++) {
+            uint8_t id = b.open_id[k];
+            if (id == h_id_html || id == h_id_head || id == h_id_body)
+                continue;
+            if (!k_allowed[id]) {
+                h_diag(&b, "expected-closing-tag-but-got-eof",
+                       b.input_len,
+                       "elements left open at end of input");
+                break;
+            }
+        }
     }
     /* Nokogiri document shape (#659): with no explicit <html>,
      * synthesize <html><head></head><body>content</body></html> —

data/vendor-src/libleptris/test/html/test_html5lib.cpp

@@ -978,8 +978,8 @@ TEST(Html5LibCorpus, ErrorVerdictParity) {
         printf("  VERDICT %s\n", failures[i].c_str());
     EXPECT_GT(total, (size_t)1500);
     /* Verdict-parity floor — set at the slice-2 baseline
-     * (1394/1555; was 190 before the #1771 emission slices —
+     * (1409/1555; was 190 before the #1771 emission slices —
      * the doctype-less initial-mode family, trailing chars, EOF with
      * open elements; slice 3 adds the raw-text EOF variants). Only raises. */
-    EXPECT_GE(parity, (size_t)1394);
+    EXPECT_GE(parity, (size_t)1409);
 }

data/vendor-src/libleptris/vcpkg.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://raw.githubusercontent.com/microsoft/vcpkg/master/scripts/vcpkg.schema.json",
   "name": "leptris",
-  "version": "1.9.374",
+  "version": "1.9.375",
   "description": "High-performance XML parser and XPath 1.0 engine in C (libleptris)",
   "homepage": "https://github.com/leptris/leptris",
   "license": "MIT",

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: leptris
 version: !ruby/object:Gem::Version
-  version: 1.9.374.0
+  version: 1.9.375.0
 platform: aarch64-linux
 authors:
 - Ribose Inc.
@@ -206,6 +206,7 @@ files:
 - vendor-src/libleptris/CONTEXT.md
 - vendor-src/libleptris/LICENSE.md
 - vendor-src/libleptris/README.adoc
+- vendor-src/libleptris/TODO.SOTA.md
 - vendor-src/libleptris/TODO.assemble/README.md
 - vendor-src/libleptris/TODO.bindings/01-mutation-construction.md
 - vendor-src/libleptris/TODO.bindings/02-incremental-iterparse.md