9.0.0 — Host objects across all engines, C++ workers, DevTools & build modernization#43
Open
ammarahm-ed wants to merge 65 commits into
Open
9.0.0 — Host objects across all engines, C++ workers, DevTools & build modernization#43ammarahm-ed wants to merge 65 commits into
ammarahm-ed wants to merge 65 commits into
Conversation
Port of upstream nativescript/android-runtime #1826 (2bab8f5). URL::New unconditionally string-read argv[1] whenever argc > 1, which threw for new URL(x, undefined) / new URL(x, null). Gate the base-URL branch on argv[1] being neither undefined nor null (napi_util::is_of_type), falling through to the no-base parse otherwise. Ports the expanded testURLImpl.js suite (engine-agnostic JS) and adds the migration ledger. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1820 (94ddb15). Use _Exit(0) instead of exit(0) on the locked-screen/direct-boot metadata path so we skip atexit handlers and destructors (a destroyed mutex during exit(0) was causing an ANR). In the napi runtime this error path lives in MetadataBuilder.cpp rather than MetadataNode.cpp. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1825 (a983931). compileSdk/targetSdk provided via -P arrive as String; cast them to int in computeCompileSdkVersion/computeTargetSdkVersion so the Android DSL accepts them instead of erroring. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1818 (3423e6f). Add -Wl,-z,max-page-size=16384 link option for arm64-v8a/x86_64 so the runtime .so is compatible with 16 KB page-size devices. Also bump default compileSdk and build-tools to 35. Skipped from the upstream commit: the gradle 8.4->8.7 wrapper and AGP 8.3.2->8.5.0 bumps (this fork is already on gradle 8.14.3 / AGP 8.12.1), and the c++_shared->c++_static STL switch (the napi multi-engine build relies on shared libc++ alongside the engine .so files). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime Ada v3 chain (#1830, #1832,
#1835, #1884: b31fc5f, 3633aed, 83f611b, 3513ce7, 45fb275). Ada is a vendored
single-file library, so the intermediate version bumps collapse into one
upgrade to the final 3.3.0.
- Upgrade vendored ada.{h,cpp} 2.9.0 -> 3.3.0 (no API breaks in URL/URLSearchParams).
- Port the URLPattern API from V8 to Node-API: new modules/url/URLPattern.{cpp,h}
backed by ada's url_pattern with a napi RegExp provider (RegExp built via the
global constructor and driven through .exec; a move-only NapiRegex wraps the
napi_ref; a thread-local env lets create_instance reach the current env). Wire
URLPattern::Init into NSRuntimeModules so it is installed as a global.
- Register the property as the spec-correct hasRegExpGroups (the original V8
impl exposed a typo'd hasRegexpGroups).
- Port testURLPattern.js and register it in mainpage.js.
Verified: native build across all ABIs (V8-13) exit 0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of the portable parts of upstream nativescript/android-runtime #1819 (bec401c, 'NDK 27 and Support for Java 21'). - NanoWSD.java: explicit (byte) casts on compound-assignment (Java 21 rejects the implicit narrowing). - NativeScriptAbstractMap: bound SimpleEntry/SimpleImmutableEntry type params with 'extends Serializable' to satisfy Java 21's serializable checks. - android-metadata-generator / runtime-binding-generator: use JavaVersion.VERSION_17 instead of the '17' string. - runtime: add explicit compileOptions VERSION_17. Skipped from upstream: CI workflows, gradlew/wrapper and root package.json version bumps (this fork is already on newer gradle/NDK), and the NDK-27 android_support/FLEXIBLE_PAGE_SIZES bits (already present here). Verified: :runtime:compileDebugJavaWithJavac under JDK 21, exit 0. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1845 (87f7f9c).
Replace the vendored libzip headers (cpp/zip/include/{zip.h,zipconf.h}) and
the four prebuilt static libs (libs/common/<abi>/libzip.a) with 1.11.4. The
only consumer, AssetExtractor.cpp, uses stable libzip API unaffected by the
bump. Verified: native compile+link across all ABIs (V8-13) exit 0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1848 (78589fd). Bump ns_default_kotlin_version / ns_default_kotlinx_metadata_jvm_version to 2.2.20 and adapt KotlinClassDescriptor to the kotlinx-metadata 2.2.x API: metaClass.enumEntries was replaced by kmEnumEntries (a list of entries), so map it to names before passing to getEnumEntriesAsFields. Verified: metadata-generator compiles under kotlin 2.2.20 and :app:buildMetadata runs it successfully. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1868 (f033061). Register a global queueMicrotask(callback). The V8 original called isolate->EnqueueMicrotask; Node-API has no equivalent, so schedule the callback via Promise.resolve().then(callback), which enqueues a real microtask on every engine the runtime targets and preserves ordering with Promise microtasks (running before timers). Throws TypeError when the argument is not a function. Ports testQueueMicrotask.js and registers it in mainpage.js. Verified: native compile across ABIs exit 0 (on-device ordering asserts not run). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1881 (5cb66ee). A weak global ref cached in the ObjectManager LRUCache can end up pointing to a collected Java object; reusing it crashes. Add an optional cache-validity callback to LRUCache that is consulted on lookup — ObjectManager supplies ValidateWeakGlobalRefCallback (IsSameObject(obj, NULL)) so stale weak refs are evicted and re-created instead of returned. Adds JEnv::isSameObject and LRUCache::evictKey. Verified: native compile across ABIs exit 0. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1886 (c05e283). Make ObjectManager::GetClassName (and the JAVA_LANG_CLASS / GET_NAME_METHOD_ID JNI ids it uses) static, so NativeScriptException can resolve a Java class name during a rethrow without going through Runtime::GetRuntime(env)->GetObjectManager(). That retrieval can throw when the env/runtime is unavailable mid-exception, which is exactly when rethrow runs. GetJavaExceptionFromEnv keeps the instance (it needs GetOrCreateObjectId/CreateJSWrapper). The runtime-lookup robustness from the upstream commit is already present in the napi GetRuntime(env). Verified: native compile+link across ABIs exit 0. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1887 (3ecd707). - Stop finalizing buildMetadata by the R8 minify task (the ordering caused proguard build failures). - Ship consumer proguard rules from the runtime library (consumerProguardFiles + consumer-rules.pro) that keep com.tns.*, generated bindings (com.tns.gen.**), internal classes, and RuntimeVisibleAnnotations so reflection/metadata survive minification. Verified: gradle config evaluation exit 0 (full effect requires a release R8 build). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of the remaining gap from upstream nativescript/android-runtime #1895 (3e61cef). The napi ForEach already used the spec callback order (value, key, searchParams), the optional thisArg, and stopped on a thrown callback. It still iterated via get_keys()+get(key), which returns the first value for a repeated key, so ?a=1&a=2 reported (1) twice. Switch to get_entries() so each entry yields its own value. Adds the forEach unit tests. Verified: native compile across ABIs exit 0. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Partial port of upstream nativescript/android-runtime #1908 (e924542). Install a std::terminate handler (LogAndAbortUncaught) in Runtime::Init that logs the in-flight native exception's message before _Exit, so an uncaught NativeScriptException/std::exception is diagnosable instead of a bare abort. Adds NativeScriptException::what() to expose the stored message. The upstream commit's larger NativeScriptException logging rewrite is V8-stack specific and does not map onto this fork's napi exception path (which already builds its own messages/stack traces); not ported. Verified: native compile across ABIs exit 0. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1920 (1fd144f). When multithreaded JS is enabled the runtime's instance maps are accessed from multiple threads, so back them with thread-safe maps. Widen the field types to the Map<> interface (only Map methods are used) and, when getEnableMultithreadedJavascript() is set, initialize them as ConcurrentHashMap / Collections.synchronizedMap. loadedJavaScriptExtends stays a synchronized HashMap because it stores null values. Adds ConcurrentAccessTest.java + testConcurrentAccess.js. Verified: :runtime: and :app: Java compile exit 0 (concurrency behavior is device-only). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Record 085bc4f + 3c956cf as deferred — a JNI perf optimization that must be ported together and device-validated across Android 8-11+ (ART calling-convention crash risk), which can't be verified in this environment. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1906 (df4e81b). In callJSMethod, re-resolve the owning runtime not just when the runtime id misses the cache but also when the resolved runtime does not actually own the target object (getJavaObjectID == null). This happens when a worker fires a JS method on an object created in the main thread or another worker. Also surface a clearer error from ObjectManager::GetJavaObjectByJsObject when GetJavaObjectByID fails (includes the object id), which usually indicates the id belongs to a different runtime/thread. Verified: native + Java compile exit 0. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…dentifier .extend keys
Batches three upstream jsparser fixes (build/ is gitignored, rebuilt on demand):
- 54185ff (#1796): webpack devtool:false (we don't use the sourcemap).
- e963d6c (#1797): traverseFiles throws when no files are found, so a broken
webpack build fails loudly instead of silently succeeding.
- dd2984b (#1950): _getIdentifierKeyName guards the .extend({}) property walk
against spreads, computed keys, and non-Identifier keys, which can't map to a
Java binding and previously crashed the parser (e.g. bundled Zod schemas).
Verified: jsparser 'npm run build' (webpack) compiles cleanly.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1910 (49da71b). Drop the buildMetadata task's dependsOn on compile*ArtProfile and optimize*Resources — under proguard/R8 these created a circular task dependency that broke the build. Verified: :app: config evaluation exit 0 (full effect requires a release R8 build). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Port of upstream nativescript/android-runtime #1938 (2b6cb03). Harden dex directory setup: create code_cache/secondary-dexes if missing and fall back to files/secondary-dexes when it is unavailable/unwritable (RuntimeHelper). Defensively create the parent directory in ProxyGenerator.saveProxy before createNewFile to avoid ENOENT crashes on newer Android versions. Verified: app + generator Java compile exit 0. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…968) Record c9d41e6 + fce8e29 as deferred — reflection into hidden BaseDexClassLoader internals, API-level-branched, with a release-only crash the second commit fixes. Must be ported as the combined end-state and validated on real devices across Android versions + a release build, which isn't possible here. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
… inspector, timers, URLSearchParams spec) Complete the 2bab8f5..HEAD triage: defer the ESM-dependent module-path fixes (92c2654/45ed1f6), the workers->C++ refactor (a84d3c7), the two inspector features (55da2da/4b5ab0a), the timers rewrite (bfd7650), and the URLSearchParams spec overhaul (89893ae), each with rationale. Add a status summary to the ledger. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…tration Port of upstream nativescript/android-runtime #1921 + #1942 (085bc4f + 3c956cf) as their combined end-state. Hot native calls (generateNewObjectId, getCurrentRuntimeId, getPointerSize as @CriticalNative; notifyGc, SetManualInstrumentationMode as @fastNative) use Android's optimized JNI calling conventions. Because dynamic JNI lookup of these annotated methods is unimplemented on Android 8-10 and buggy on 11, JNI_OnLoad binds the annotated variants via RegisterNatives, and the Java side dispatches to auto-bound *Legacy variants when SDK_INT < 26. generateNewObjectId becomes static and is called with (nullptr, nullptr) since its impl ignores env/obj; getCurrentRuntimeId uses Runtime::Current() so it needs no JNIEnv. notifyGc keeps this fork's (int, int[]) signature -> "(I[I)Z". Verified: native + Java compile across ABIs exit 0. Calling-convention correctness must be validated on-device (Android 8-11, 12+, and a release build). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Collaborator
|
@ammarahm-ed awesome. Will only be able to test in two weeks |
Collaborator
Author
Doesn't work still on nativescript 9.x, all runtime tests pass yet ns 9 project doesn't work. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Uh oh!
There was an error while loading. Please reload this page.