Remove the lib.name documentation lie + dead LlamaSystemProperties.getLibName()

Deep forensic comparison against kherud/java-llama.cpp (the upstream)
showed that net.ladenthin.llama.lib.name has been a documentation-only
ghost since 2023:

Upstream history of "lib.name":
  c8f57f7 (initial JNI commit) - Javadoc mention AND real code consumer:
    String nativeLibName = System.getProperty("de.kherud.llama.lib.name");
    if (nativeLibName == null) { ... }
  7a852d1 - README documents it (-Dde.kherud.llama.lib.name=myname.so)
  6bb63e1 ("add ggml shared library to binding") - CODE CONSUMER DELETED.
    The loader was extended to load multiple shared libraries (ggml and
    jllama as separate files); the single-name-override model became
    incompatible. Javadoc + README mentions were left behind.
  6d0c4af - README mention removed.

  The Javadoc lie in LlamaLoader.java survives upstream to this day.

Bernard's fork inherited the Javadoc lie verbatim and later added a
LlamaSystemProperties.getLibName() getter for "completeness" (every
documented property gets a registry getter) - but the loader code still
hardcodes "jllama" and ignores the override.

Why delete rather than re-wire:
- Upstream deliberately removed the consumer to support multi-library
  loading. Re-wiring would either reintroduce the same single-name
  limitation upstream removed, or need a new design covering both
  libraries.
- This fork has added even more dimensions (CPU / CUDA /
  OpenCL-Adreno classifiers); the "one override filename" idea is
  even less applicable now than in upstream's day.
- No caller has ever asked for it; the README will be more honest
  with one fewer documented-but-broken row.

Changes:
- LlamaSystemProperties.getLibName() deleted (~5 lines).
- LlamaLoader.java Javadoc fixed: lib.name removed from the "set these
  properties" sentence; replaced with a multi-paragraph history note
  citing upstream commit 6bb63e1 so future readers understand the
  property's lifecycle and don't try to "re-add" it.
- README System Properties Reference: lib.name row dropped (was 9 rows,
  now 8).
- CLAUDE.md: the registry-cleanup TODO is closed for both subitems
  (lib.name DELETED here; osinfo.architecture FIXED in 3ae6c81 by
  routing OSInfo.getArchName() through the registry getter).

Tests: OSInfoTest 16/16 + LlamaLoaderTest 21/21 = 37/37 pass.
mvn compile clean.

Author: claude

CLAUDE.md

@@ -688,9 +688,9 @@ See [`../workspace/policies/jqwik-prompt-injection.md`](../workspace/policies/jq
 
 - **Cross-repo code-quality TODOs** — see [`../workspace/policies/code-quality-todos.md`](../workspace/policies/code-quality-todos.md) for the canonical `@VisibleForTesting` design-fit review, package hierarchy review, and class/method naming review. This repo has no `@VisibleForTesting` usages today; package and naming reviews remain open.
 
-- **`LlamaSystemProperties` registry cleanup (deep-scan finding).** The deep scan that produced the README [System Properties Reference](README.md#system-properties-reference) surfaced two small inconsistencies between the registry class and its actual consumers — worth a focused cleanup but low priority:
-  - `LlamaSystemProperties.getLibName()` is **declared but has zero callers** in `src/main/java`. Either wire it into `LlamaLoader`'s filename-resolution path (so `-Dnet.ladenthin.llama.lib.name=…` actually does something), or delete the getter as dead code. The README table documents the property as "currently no production caller" so users are not misled in the meantime.
-  - `OSInfo.java:390` reads `System.getProperty("net.ladenthin.llama.osinfo.architecture")` directly with the literal string, **bypassing** `LlamaSystemProperties.getOsinfoArchitecture()`. The duplication parallels the recent BAF `Radix.HEX` consolidation: the registry-side getter exists but is not the single source of truth. Route `OSInfo` through the registry getter so a future property rename only has to land in one place.
+- ~~**`LlamaSystemProperties` registry cleanup (deep-scan finding).**~~ ✅ **BOTH RESOLVED** (this session). The deep scan that produced the README [System Properties Reference](README.md#system-properties-reference) surfaced two registry-bypass smells; both have now been fixed:
+  - ~~`LlamaSystemProperties.getLibName()` is declared but has zero callers~~ ✅ **DELETED**. Forensic trace against `kherud/java-llama.cpp` history (cloned into `/tmp` this session) showed the loader code that originally read `lib.name` was removed in upstream commit `6bb63e1` (*"add ggml shared library to binding"*) when the loader was extended to load multiple shared libraries (ggml + jllama) as separate files — the single-name-override model became incompatible. The Javadoc + README mentions of `lib.name` lived on in both upstream and this fork as a documentation lie ever since. Re-wiring it would either reintroduce the one-library limitation upstream removed, or need a much bigger multi-library design with no caller asking for it. Cleanup landed: getter deleted from `LlamaSystemProperties`, Javadoc lie removed from `LlamaLoader` (with a comment block citing `6bb63e1` so future readers know the history), README row dropped.
+  - ~~`OSInfo.java:390` bypasses `LlamaSystemProperties.getOsinfoArchitecture()`~~ ✅ **FIXED** in commit `3ae6c81` — `OSInfo.getArchName()` now routes through `new LlamaSystemProperties().getOsinfoArchitecture()`; the literal property string lives in exactly one place (the registry). Direct parallel to the recent BAF `Radix.HEX` consolidation.
 
 - ~~**Abstract the Java and test writing guidelines to a workspace-level shared layer.**~~ **DONE.** This repo is Java 8; follow the workspace version chain at [`../workspace/guides/src/CODE_WRITING_GUIDE-8.md`](../workspace/guides/src/CODE_WRITING_GUIDE-8.md) and [`../workspace/guides/test/TEST_WRITING_GUIDE-8.md`](../workspace/guides/test/TEST_WRITING_GUIDE-8.md). Canonical TDD skill at [`../workspace/.claude/skills/java-tdd-guide/SKILL.md`](../workspace/.claude/skills/java-tdd-guide/SKILL.md). This repo has no project-specific writing-guide supplements.
 

README.md

@@ -254,7 +254,6 @@ Every `net.ladenthin.llama.*` system property recognised by the library, deep-sc
 | Property | Default | Scope | Consumer | Description |
 |---|---|---|---|---|
 | `net.ladenthin.llama.lib.path` | unset (falls back to `java.library.path`) | runtime | `LlamaLoader` | Directory containing the native `jllama` shared library. Checked first, before `java.library.path`. Set with `-Dnet.ladenthin.llama.lib.path=/path/to/dir`. |
-| `net.ladenthin.llama.lib.name` | unset (platform-determined, e.g. `jllama.so`) | runtime | `LlamaSystemProperties.getLibName()` (declared, currently no production caller) | Override for the native library filename. |
 | `net.ladenthin.llama.tmpdir` | unset (falls back to `java.io.tmpdir`) | runtime | `LlamaLoader` | Custom temporary directory used when extracting the native library from the JAR. |
 | `net.ladenthin.llama.osinfo.architecture` | unset (uses `os.arch`) | runtime | `OSInfo` | Override for the architecture string used to locate the bundled library inside the JAR. Useful when `os.arch` reports an unexpected value (e.g. inside dockcross / chrooted environments). |
 | `net.ladenthin.llama.test.ngl` | `43` | test | `LlamaModelTest`, `RerankingModelTest`, `ChatScenarioTest`, `ChatAdvancedTest`, `ErrorHandlingTest`, `SessionConcurrencyTest`, `ConfigureParallelInferenceTest`, `MultimodalIntegrationTest` (via `Integer.getInteger(TestConstants.PROP_TEST_NGL, TestConstants.DEFAULT_TEST_NGL)`) | Number of GPU layers used during testing. Pin to `0` on CPU-only hosts: `mvn test -Dnet.ladenthin.llama.test.ngl=0`. |

src/main/java/net/ladenthin/llama/LlamaLoader.java

@@ -19,12 +19,22 @@
 import org.jspecify.annotations.Nullable;
 
 /**
- * Set the system properties {@code net.ladenthin.llama.lib.path} /
- * {@code net.ladenthin.llama.lib.name} appropriately so that the library can
- * find *.dll, *.dylib and *.so files, according to the current OS (win, linux, mac).
+ * Set the system property {@code net.ladenthin.llama.lib.path} appropriately
+ * so that the library can find {@code *.dll}, {@code *.dylib} and
+ * {@code *.so} files, according to the current OS (Windows, Linux, macOS).
  *
  * <p>The library files are automatically extracted from this project's package (JAR).
  *
+ * <p>Historically the loader also honoured a {@code net.ladenthin.llama.lib.name}
+ * property that overrode the resolved library filename. Upstream removed the
+ * code path that read it in {@code kherud/java-llama.cpp} commit {@code 6bb63e1}
+ * (&quot;add ggml shared library to binding&quot;) when the loader was extended to
+ * load multiple shared libraries (ggml + jllama) as separate files &mdash; the
+ * single-name-override model is incompatible with that. The Javadoc mention
+ * has since been a documentation lie in both upstream and this fork; it has
+ * now been removed here, and the corresponding {@code getLibName()} getter
+ * has been deleted from {@code LlamaSystemProperties}.
+ *
  * <p>usage: call {@link #initialize()} before using the library.
  *
  * @author leo

src/main/java/net/ladenthin/llama/LlamaSystemProperties.java

@@ -31,15 +31,6 @@ public LlamaSystemProperties() {}
         return getProperty(".lib.path");
     }
 
-    /**
-     * Override for the native library file name.
-     *
-     * @return the configured library file name, or {@code null} if unset
-     */
-    public @Nullable String getLibName() {
-        return getProperty(".lib.name");
-    }
-
     /**
      * Custom temporary directory used when extracting the native library from
      * the JAR. Falls back to {@code java.io.tmpdir} if absent.