Merge pull request #964 from fglock/fix/jcpan-compiler-tooling-modules-20260815

Fix CPAN compiler, runtime, and jcpan tooling blockers

Author: fglock

build.gradle

@@ -220,6 +220,7 @@ dependencies {
     implementation libs.jing                         // RELAX NG validation for XML::LibXML
     implementation libs.snakeyaml.engine              // YAML processing
     implementation libs.tomlj                         // TOML processing
+    implementation libs.zxing.core                    // QR encoding for Text::QRCode
     implementation libs.commons.csv                   // CSV processing
     implementation libs.commonmark                    // CommonMark rendering
     implementation libs.commonmark.autolink           // GFM autolinks

dev/design/jcpan-compiler-tooling-followup.md

@@ -37,7 +37,7 @@ The following are not treated as PerlOnJava regressions because their current di
 
 ## Progress Tracking
 
-### Current Status: implementation complete; PR ready for review
+### Current Status: unified into PR #964; combined CI passing
 
 ### Completed Phases
 
@@ -59,11 +59,34 @@ The following are not treated as PerlOnJava regressions because their current di
   - Opened PR #962 from `fix/jcpan-compiler-tooling-followup`.
   - GitHub Actions passed on Ubuntu and Windows; the Windows run specifically
     confirmed the platform-default newline fix in `pipe_jperl_shebang.t`.
+- [x] Phase 7: post-merge warning cleanup (2026-08-15)
+  - Replaced deprecated `Zstd.decompressedSize` calls with
+    `Zstd.getFrameContentSize` and reject unknown or invalid frame sizes before
+    allocating decoder buffers.
+- [x] Phase 8: core-suite regression audit (2026-08-15)
+  - Compared the reported core files against isolated current-master and
+    historical baseline builds instead of treating aggregate TAP counts as
+    deterministic.
+  - Fixed PVLV filehandle handling so `-t` follows the glob's IO slot without
+    stringification and `close` warnings retain ASCII and Unicode glob names.
+  - Restored `op/gv.t` from 253/304 to 255/304 and `uni/gv.t` from 176/206 to
+    178/206. `re/pat_advanced.t`, `re/pat_advanced_thr.t`, and
+    `test_pl/examples.t` reproduce their higher reported counts. The remaining
+    `japh/abigail.t` 109/130 result matches both current master and the
+    documented historical baseline, so it is not a PR #963 regression.
+  - Added a four-case system-Perl oracle and verified it with both PerlOnJava
+    backends; the full `make` suite passes.
+  - PR #963 CI passed on Ubuntu and Windows at commit `b48f504d8`.
+- [x] Phase 9: unified review branch (2026-08-15)
+  - Merged all five PR #963 commits into PR #964 so the compiler, runtime,
+    CPAN-tooling, and Java-module changes can be tested and approved together.
+  - The combined full `make` suite passed before the unified branch was pushed.
+  - Unified PR #964 CI passed on Ubuntu in 14m35s and Windows in 17m27s.
 
 ### Next Steps
 
-1. Review PR #962.
-2. Merge after approval.
+1. Mark PR #963 as superseded by PR #964.
+2. Hand PR #964 back for user testing and review.
 
 ### Open Questions
 

dev/design/jcpan-compiler-tooling-modules.md

@@ -0,0 +1,66 @@
+# jcpan compiler and tooling modules
+
+## Goal
+
+Fix reusable compiler and CPAN-tooling blockers exposed by Pegex::JSON,
+Music::Factory, Bio::Data::Plasmid::CloningVector,
+Catalyst::Engine::HTTP::POE::YieldCC, App::Chained, Queue, and their
+dependencies. Prefer shared runtime/tooling fixes and existing Java libraries
+over distribution preferences.
+
+## Progress Tracking
+
+### Current Status: PR #963 commits unified into PR #964; combined CI passing
+
+### Completed Phases
+
+- [x] Phase 1: baseline and system-Perl classification (2026-08-15)
+  - Captured bounded `jcpan -t` logs for all six requested modules.
+  - Identified malformed Bio-MCPrimers packaging and Catalyst's omitted
+    Restarter::Watcher as upstream distribution failures.
+- [x] Phase 2: shared root-cause implementation (2026-08-15)
+  - Restored YAML::PP's standard object `dump` API.
+  - Serialized scalar references through their referents for boolean.pm parity.
+  - Made gzip EOF status compatible with CPAN single-file extraction.
+  - Extended generic missing-prerequisite discovery to TAP diagnostics.
+  - Routed Object::Pad's core syntax to PerlOnJava's native class compiler.
+- [x] Phase 3: cross-runtime regression validation (2026-08-15)
+  - Validated YAML, gzip, native-class, and CPAN-tooling regressions with
+    system Perl where applicable.
+  - Passed the focused regressions on JVM and interpreter backends.
+- [x] Phase 4: requested module verification (2026-08-15)
+  - Pegex::JSON: 4 files, 21 assertions, PASS.
+  - Music::Factory: 5 files, 20 assertions, PASS.
+  - App::Chained: 2 files, 9 assertions, PASS after generic dependency retry.
+  - Queue: single-file distribution built and tested successfully; upstream
+    ships no test directory.
+  - Bio::Data::Plasmid::CloningVector excluded because Bio-MCPrimers has no
+    Makefile.PL and fails system-Perl configuration.
+  - Catalyst::Engine::HTTP::POE::YieldCC excluded because its distribution
+    requires but does not ship or declare Restarter::Watcher; system Perl
+    reproduces the missing-module failure.
+- [x] Phase 5: full verification (2026-08-15)
+  - Full `make` passed all unit shards.
+- [x] Phase 6: pull request and CI (2026-08-15)
+  - Opened [PR #964](https://github.com/fglock/PerlOnJava/pull/964).
+  - Ubuntu CI passed in 11m20s.
+  - Windows CI passed in 18m10s.
+- [x] Phase 7: unified review branch (2026-08-15)
+  - Merged all five commits from PR #963 into PR #964 for joint testing and
+    approval.
+  - The combined full `make` suite passed before push.
+  - Unified PR #964 CI passed on Ubuntu in 14m35s and Windows in 17m27s.
+
+### Next Steps
+
+1. Mark PR #963 as superseded by PR #964.
+2. Await joint testing and review of PR #964.
+
+### Open Questions
+
+- Object::Pad-specific MOP and extension APIs remain outside the native class
+  compatibility pragma; the requested Music::Factory surface uses core syntax.
+
+## References
+
+- Skills: `debug-perlonjava`, `port-cpan-module`

docs/about/changelog.md

@@ -4,6 +4,11 @@ Release history of PerlOnJava. See [Roadmap](roadmap.md) for future plans.
 
 ## Work in progress
 
+- CPAN/compiler tooling: restore YAML::PP's object `dump` API, recognize
+  TAP-indented missing prerequisites, report gzip stream completion for CPAN
+  single-file distributions, and route Object::Pad's core syntax through the
+  native class compiler. This unblocks Pegex::JSON, Music::Factory,
+  App::Chained, and Queue without distribution preferences.
 - CPAN/compiler tooling: add transitive prerequisites to the bundled-provider
   manifest, provide a JAXP-backed `XML::LibXSLT`, and preserve descriptors for
   anonymous handles stored in container lvalues. This unblocks

docs/reference/bundled-modules.md

@@ -377,6 +377,7 @@ These are loaded automatically or via `use`:
 
 | Module | Implementation | Notes |
 |--------|---------------|-------|
+| `Object::Pad` | Perl/compiler | Core `class`, `field`, `method`, `:param`, and `:isa` syntax uses PerlOnJava's native class compiler; Object::Pad MOP extensions are not included |
 | `Scalar::Util` | Java | |
 | `Sub::Name` | Java | |
 | `Sub::Util` | Java | |

docs/reference/feature-matrix.md

@@ -771,6 +771,9 @@ The `:encoding()` layer supports all encodings provided by Java's `Charset.forNa
 - ❌  **Safe** module.
 
 ### Non-core modules
+- 🟡 **Object::Pad**: core class, field, method, parameter, and inheritance
+  syntax is handled by PerlOnJava's native class compiler; Object::Pad-specific
+  MOP extensions are not implemented.
 - ✅  **JSON::DWIW**: relaxed JSON conversion implemented over the bundled
   pure-Perl `JSON::PP` backend.
 - ✅  **Taint::Runtime**: Java XS replacement for runtime taint toggling and

docs/reference/xs-compatibility.md

@@ -26,6 +26,8 @@ These modules have optimized Java implementations built into PerlOnJava:
 | Exporter::Lexical | ExporterLexical.java | 0.02 | Installs lexical subs into the enclosing compile-time scope |
 | HTML::Content::Extractor | HTMLContentExtractor.java | 0.17 | Uses jsoup's HTML5 parser with a small legacy tree-compatibility layer |
 | Crypt::Twofish2 | CryptTwofish2.java | 1.03 | Uses BouncyCastle Twofish; ECB, stateful zero-IV CBC, and CFB1 |
+| Math::Cephes | MathCephes.java | 0.5308 | Implements the normal and chi-square distribution functions used by CPAN statistics modules |
+| Text::QRCode | TextQRCode.java | 0.05 | Uses ZXing in forced byte mode with libqrencode-compatible mask selection |
 
 ## Modules with PP Fallbacks or Shims
 

gradle/libs.versions.toml

@@ -13,6 +13,7 @@ snappy-java = "1.1.10.8"
 sqlite-jdbc = "3.53.2.1"
 tomlj = "1.1.1"
 zstd-jni = "1.5.7-8"
+zxing = "3.5.4"
 
 [libraries]
 asm = { module = "org.ow2.asm:asm", version.ref = "asm" }
@@ -37,6 +38,7 @@ snappy-java = { module = "org.xerial.snappy:snappy-java", version.ref = "snappy-
 sqlite-jdbc = { module = "org.xerial:sqlite-jdbc", version.ref = "sqlite-jdbc" }
 tomlj = { module = "org.tomlj:tomlj", version.ref = "tomlj" }
 zstd-jni = { module = "com.github.luben:zstd-jni", version.ref = "zstd-jni" }
+zxing-core = { module = "com.google.zxing:core", version.ref = "zxing" }
 
 [plugins]
 cyclonedx = "org.cyclonedx.bom:2.3.0"

jcpan

@@ -116,7 +116,7 @@ export JPERL_ORPHAN_EXIT=1
 # CPAN build tools may install copies of their own implementation modules into
 # the user library.  Keep PerlOnJava's narrow compatibility overlays ahead of
 # those copies while jcpan and its child build processes run.
-export PERLONJAVA_PREFER_BUNDLED_MODULES="Module/Build/Base.pm${PERLONJAVA_PREFER_BUNDLED_MODULES:+,$PERLONJAVA_PREFER_BUNDLED_MODULES}"
+export PERLONJAVA_PREFER_BUNDLED_MODULES="Module/Build/Base.pm,Object/Pad.pm${PERLONJAVA_PREFER_BUNDLED_MODULES:+,$PERLONJAVA_PREFER_BUNDLED_MODULES}"
 
 # CPAN test suites should run with deterministic semantics. User-interface
 # color preferences such as NO_COLOR can change module behavior under test

src/main/java/com/booking/sereal/Decoder.java

@@ -334,7 +334,13 @@ private void uncompressZstd() throws SerealException {
     int len = (int) read_varint();
 
     byte[] compressedData = Arrays.copyOfRange(originalData.array, position, position + len);
-    long decompressedSize = Zstd.decompressedSize(compressedData);
+    long decompressedSize = Zstd.getFrameContentSize(compressedData);
+    if (Zstd.isError(decompressedSize)) {
+      String message = decompressedSize == -1
+          ? "Zstd frame content size is unknown"
+          : Zstd.getErrorName(decompressedSize);
+      throw new SerealException(message);
+    }
 
     if (decompressedSize > this.maxSize) {
       throw new SerealException("The expected uncompressed size is larger than the allowed maximum size");