The session-long set of changes that lay the groundwork for the
Jai-literal implicit-Context-parameter refactor. Lots of accumulated
work; the new arrival is the implicit-ctx foundation (steps 1+2 of
the plan in current/CHECKPOINT-MEM.md):
Step 1 — `CAllocator :: struct {}` stateless allocator in
library/modules/allocators.sx, delegating directly to
libc_malloc/libc_free. `ConstantValue` in src/ir/inst.zig gains a
`func_ref: FuncId` leaf so nested aggregates can carry function
pointers (the inline Allocator value's fn-ptr fields). Switch
sites updated in emit_llvm.zig, print.zig, interp.zig.
Step 2 — `emitDefaultContextGlobal` in src/ir/lower.zig synthesises
a static `__sx_default_context` global with a nested-aggregate
init_val pointing at the CAllocator → Allocator thunks. The
second-pass `initVtableGlobals` in emit_llvm.zig is generalised
to handle `.aggregate` init_vals (re-emits after func_map is
populated so func_ref leaves resolve to real symbols).
Also folded in from earlier work this session:
- Phase 1.1: `xx value` heap-copy in `buildProtocolValue` routes
through `context.allocator` via the new `allocViaContext` helper.
- interp.zig: `marshalForeignArg` double-offset bug fixed —
`heapSlice` already adds `hp.offset` to the slice ptr, so the
extra `+ hp.offset` was scribbling memcpy/memset into adjacent
heap state, corrupting `heap.items[0]`. Symptom: `build_format`
at comptime produced zero bytes, all `print` calls failed.
- Lazy lowering: `lazyLowerFunction` now declares foreign-body
functions as extern stubs in the local (comptime) module so
cross-module foreign calls resolve.
- Allocator API: all stdlib allocators on one-line `init() -> *T`
(CAllocator/GPA: libc-backed; Arena/TrackingAllocator: parent-
backed; BufAlloc: embeds state at head of user buffer).
- issues 0038 (transitive #import), 0039 (chess + stdlib migration
fallout), 0040 (generic struct method dot-dispatch), 0041
(pointer types as type-arg), 0042 (alias name resolution) — all
fixed; regression tests in examples/.
- Diagnostic: `emitError` now embeds the lowering's
`current_source_file` and enclosing function in the literal
message; SX_TRACE_UNRESOLVED=1 dumps a Zig stack trace at the
emit site so misattributed spans can't hide where the failure
is.
- tools/verify-step.sh (all-platforms gate) and tools/scratch.sh
(interp/codegen parity tester) added.
Test suite: 152 example tests pass; chess builds + screenshots on
macOS / iOS sim / Android.
Week 7 of /Users/agra/.claude/plans/lets-plan-to-move-splendid-pumpkin.md
plus the android.sx refactor + three sx-compiler fixes hit along the way
to get chess on Pixel 7 Pro responding to touch end-to-end.
library/modules/platform/bundle.sx now covers the Android APK shape
alongside macOS / iOS-sim / iOS-device. `android_bundle_main` discovers
the SDK ($ANDROID_HOME / $ANDROID_SDK_ROOT / $HOME/Library/Android/sdk),
picks the highest-versioned build-tools + platforms via
`process.run("ls .. | sort -V | tail -1")`, stages
`<apk>.stage/lib/arm64-v8a/<libfoo.so>`, synthesizes
AndroidManifest.xml (NativeActivity vs `#jni_main` Activity branch),
writes each `#jni_main` decl's Java source under
`<stage>/java/<pkg>/<Cls>.java`, runs javac --release 11 + d8 to
produce classes.dex, aapt2-links the unaligned APK, appends lib/ +
classes.dex + each registered asset tree via zip, zipalign + ensure
debug keystore via keytool + apksigner sign.
Compiler-side accessors (src/ir/compiler_hooks.zig + library/modules/compiler.sx):
- is_android predicate.
- set_manifest_path / manifest_path + set_keystore_path / keystore_path.
- jni_main_count / jni_main_foreign_path_at(i) /
jni_main_java_source_at(i) surface the `#jni_main` emissions that
the Zig createApk previously consumed directly.
- main.zig wires manifest_path, keystore_path, and the per-decl
(foreign_path, java_source) parallel slices into BuildConfig before
invoking the post-link callback.
CLI `--apk <path>` keeps working as a transitional alias: it now feeds
bundle_path so the existing auto-`post_link_module = "platform.bundle"`
shim fires the same way as `--bundle`. main.zig no longer calls
target.createApk directly.
Deletions in src/target.zig: createApk, compileJniMainSources,
buildJniMainManifest, buildAndroidManifest, ensureDebugKeystore,
libNameFromSoBasename, plus helpers splitForeignPath / discoverJavac /
discoverAndroidSdk / findHighestSubdir / runProcess / runProcessIn
(~400 lines). git grep returns only the obituary comment.
library/modules/platform/android.sx refactor (chess Android dependency):
- Module-level globals retired (g_app_window, g_egl_*, g_viewport_*,
g_dpi_scale, g_should_stop, g_render_thread*, g_user_main_fn,
g_touch_*) → AndroidPlatform struct fields.
- All sx_android_* helpers take `plat: *AndroidPlatform` as first arg.
Render thread receives plat via pthread_create's arg.
- New `logical_w: f32 = 0.0` field. Consumers set it before init() to
define the design width in points; `recompute_scale` derives
`dpi_scale = pixel_w / logical_w` (or 1.0 if unset). Called on
init / set_viewport / egl_init. drain_touches divides incoming
physical pixel coords by dpi_scale so chess sees logical-space
positions matching its layout. Touch lands on the right squares.
Three sx-compiler bugs hit + fixed along the way:
1. Top-level `inline if OS == .X { decls }` body decls were silently
dropped because scanDecls/lowerDecls had no .if_expr arm. New
`flattenComptimeConditionals` pre-pass in src/imports.zig
(threaded via ComptimeContext from core.zig) hoists matching arms
recursively. Regression at examples/124-inline-if-hoist-toplevel.sx.
2. Parser rejected `#import` / `#framework` inside inline-if bodies
because parseStmt in src/parser.zig only had arms for `#insert`.
Added the missing arms. Regression at
examples/123-inline-if-import-in-body.sx (landed earlier).
3. JNI `Call<T>Method` switches in src/ir/emit_llvm.zig (instance /
nonvirtual / static) were missing `.f32` rows — jfloat returns
(e.g. MotionEvent.getX/getY) fell into the silent-undef else arm.
Chess's sx_android_push_touch(plat, getAction(), getX(), getY())
delivered garbage f32 coords to the touch ring, so taps landed
nowhere recognisable. Added `.f32 => Jni.Call{Static,Nonvirtual,}FloatMethod`
rows to all three switches; lifted unsupported-type detection
from emit_llvm into lowerForeignMethodCall with proper
source-spanned diagnostics (`isJniReturnTypeSupported`). Regressions
at examples/ffi-jni-call-10-jfloat-return.sx,
examples/ffi-jni-class-09-multi-float-args.sx,
examples/ffi-jni-call-11-unsupported-return-diag.sx.
Stale-snapshot drift in tests/expected/ffi-objc-call-03-selector-sharing.ir
and ffi-objc-call-06-sret-return.ir picks up the new BuildOptions
accessor extern decls (is_android, set_manifest_path,
set_keystore_path, jni_main_count, jni_main_foreign_path_at,
jni_main_java_source_at). Verified diff is dead-decl-only.
Chess on Pixel 7 Pro: tap on e2 white pawn -> yellow selection +
green dots on legal e3/e4 targets; tap on e4 -> board updates with
1. e4, "Black to move" + "1. e4" in info panel.
zig build && zig build test && bash tests/run_examples.sh -> 145/145
green. bash tests/cross_compile.sh -> 7/7 green.
Campaign Weeks 3-6 of /Users/agra/.claude/plans/lets-plan-to-move-splendid-pumpkin.md
land in one push: the bundling pipeline that used to live in
src/target.zig (createBundle, embedFramework, extractEntitlements,
buildInfoPlist, codesign) now lives in
library/modules/platform/bundle.sx and runs in the IR interpreter
after target.link() returns.
New language-side surface:
- library/modules/fs.sx — POSIX libc bindings (open/read/write/close,
mkdir/unlink/rmdir, chmod, rename, access, basename/dirname). Variadic
open() lowers to C's varargs via the new args: ..T form. Direct libc
calls bypass *File method dispatch so they work from the post-link
IR interpreter.
- library/modules/process.sx — popen-based run(cmd) returning
ProcessResult{ exit_code, stdout }, plus env() and find_executable().
- library/modules/std.sx — xml_escape(s) and variadic path_join(parts).
- library/modules/compiler.sx — BuildOptions grows
set_post_link_callback / set_post_link_module / binary_path
accessors; bundle_path/bundle_id/codesign_identity/provisioning_profile
setters + accessors; per-target predicates is_macos/is_ios/
is_ios_device/is_ios_simulator + target_triple; framework_count /
framework_at(i) / framework_path_count / framework_path_at(i);
add_asset_dir(src, dest) + asset_dir_count / src_at / dest_at.
Compiler-side wiring:
- src/ir/compiler_hooks.zig — BuildConfig now carries post_link_callback_fn,
post_link_module, binary_path, bundle_*, target_triple,
target_frameworks, target_framework_paths, asset_dirs. Hook registry
exposes every accessor; getters return "" / 0 for unset fields so
bundle.sx can treat absent values uniformly.
- src/ir/host_ffi.zig (new) — dlsym(RTLD_DEFAULT) + arity-switched cdecl
trampolines so #foreign("c") declarations resolve through the host
libc during #run / post-link interpretation.
- src/ir/interp.zig — callForeign dispatch; build_config pointer
injection so accessor hooks see live state during re-entry.
- src/core.zig — keeps the IR module alive past generateCode; exposes
invokeByName / invokeByFuncId so main.zig can re-enter the
interpreter after linking.
- src/main.zig — wires bundle/codesign/provisioning CLI flags +
target_triple + framework lists into BuildConfig; invokes the
post-link callback (by FuncId or by <module>.bundle_main lookup) once
target.link() returns. When --bundle is set but no callback is
registered, auto-falls-back to post_link_module = "platform.bundle"
so the legacy --bundle CLI keeps working for any program that imports
modules/platform/bundle.sx.
Apple .app bundler (library/modules/platform/bundle.sx):
- Single bundle_main entry covers macOS, iOS simulator, iOS device.
Per-target Info.plist switch keys off is_ios()/is_ios_simulator() —
iOS emits UIDeviceFamily / LSRequiresIPhoneOS /
UIApplicationSceneManifest / DTPlatformName (iPhoneOS or
iPhoneSimulator); macOS emits the minimal CFBundle* set.
- iOS-only steps:
- Provisioning embed: fs.read_file + fs.write_file to
<bundle>/embedded.mobileprovision.
- Framework embed: recursive cp -R per -F search path into
<bundle>/Frameworks/<Name>.framework/ (until fs.sx grows list_dir).
- Entitlements extraction: four process.run calls (security cms -D,
plutil -extract Entitlements xml1, plutil -extract
ApplicationIdentifierPrefix.0, plutil -replace application-identifier)
resolving the wildcard <TEAM>.* -> <TEAM>.<bundle_id>.
- Real codesign with --entitlements when present.
- Asset dirs (add_asset_dir): recursive cp -R src/. into <bundle>/dest/.
Missing src is treated as "nothing to do" so projects can register
add_asset_dir("assets", "assets") unconditionally.
Parser:
- parseStmt() now accepts #import \"path\"; and #framework \"Name\"; as
statement-position tokens. Needed for top-level
inline if OS == .android { #import \"modules/platform/android.sx\"; }
blocks (issue-0042 flatten pass surfaces them); chess's
inline-if-with-#import was rejected at parse time before this fix.
Removals from src/target.zig:
- createBundle, embedFramework, extractEntitlements, buildInfoPlist,
codesign (~210 lines). main.zig no longer calls createBundle after
link(); the sx callback is the single entry point.
Tests / regression markers (all run under sx run host JIT):
- examples/115-post-link-callback.sx — callback registration round-trip.
- examples/116-fs-roundtrip.sx — fs.write_file -> fs.read_file -> exists.
- examples/117-process-roundtrip.sx — process.run + env + find_executable.
- examples/118-macos-bundle.sx — macOS .app via bundle_main callback.
- examples/119-interp-cast-ptr-cmp.sx — cast(T) val under interpreter.
- examples/120-interp-variadic-any.sx — variadic ..Any indexing in IR
interpreter.
- examples/121-ios-sim-bundle.sx — iOS-sim cross-compile + .app with
iOS-shaped Info.plist (added to tests/cross_compile.sh as the
ios-sim tuple).
- examples/122-ios-device-bundle.sx — iOS device cross-compile +
full codesign pipeline (provisioning embed + entitlements
extraction + --entitlements codesign). Manually verified end-to-end:
installed via xcrun devicectl device install app + launched
successfully on iPhone 17 Pro.
- examples/123-inline-if-import-in-body.sx — locks in the parser fix.
zig build && zig build test && bash tests/run_examples.sh => 141 passed,
0 failed; bash tests/cross_compile.sh => 7 passed, 0 failed.
Trailing `args: ..T` on a #foreign declaration now lowers to the C
calling convention's `...` instead of sx-side slice-packing. Drops
the per-arity #foreign-shim workaround for callers of variadic C
APIs (__android_log_print, printf-family, etc.). Closes issue-0043.
- IR: Function.is_variadic on inst.Function; declareFunction drops
the variadic param from the IR signature for foreign+variadic
decls.
- emit_llvm: LLVMFunctionType receives is_var_arg=1 when the flag
is set; call lowering passes extras through unchanged.
- Lowering: packVariadicCallArgs early-outs for foreign+variadic
(no slice-pack); new promoteCVariadicArgs applies C default
argument promotion (bool/s8/s16/u8/u16 -> s32, f32 -> f64) to
extras past the fixed param count.
- Test: examples/ffi-foreign-cvariadic.sx + .c exercise s64/f64/s32
returns through C va_arg over s32/f64/*u8 element types.
134 host + 6 cross tests pass on the WIP-less baseline.
Adds the constructor-invocation arm of the foreign-class DSL:
`SurfaceView.new(ctx)` (where `SurfaceView` is a `#foreign #jni_class`
with `static new :: (ctx: *Context) -> *Self;`) lowers to
`FindClass(env, "android/view/SurfaceView") + GetMethodID(env, cls,
"<init>", "(args)V") + NewObject(env, cls, mid, args...)`. Returns
the fresh jobject.
- inst.zig: `JniMsgSend.is_constructor` flag + `parent_class_path`
re-purposed to carry the class being constructed (alongside its
existing nonvirtual-super-class use). Mutually exclusive with
`is_static` / `is_nonvirtual`.
- lower.zig: `lowerCall.field_access` arm now recognises
`Alias.method(args)` where `Alias` resolves in `foreign_class_map`
and the matching member is `static`. `new` routes to a new
`lowerForeignStaticCall` that derives a `(args)V` JNI descriptor
and emits a `JniMsgSend` with `is_constructor=true`. Non-`new`
static calls report a clear "use #jni_static_call" diagnostic
until that sugar lands.
- emit_llvm.zig: new `NewObject` vtable slot (28) + `emitJniConstructor`
helper expanding the FindClass+GetMethodID+NewObject chain. The
jni_msg_send arm short-circuits to it when `is_constructor` is set.
Smoke `ffi-jni-main-03-ctor.sx` exercises both this slice and the
previous super-dispatch slice in a single `onCreate` body: calls
`super.onCreate(b)` then constructs a `SurfaceView` with the Activity
as Context. IR shows the expected six-stage chain (FindClass+GetMethodID+
CallNonvirtual + FindClass+GetMethodID+NewObject); APK builds clean.
Naming caveat: the Java type `android.content.Context` clashes with
sx stdlib's `Context :: struct {...}` (heap-context). The smoke aliases
it `JContext` — future work could add a path-prefix or `as` rename
form on `#jni_class` to avoid the manual rename.
133 host / 6 cross / zig build test all green.
Inside a `#jni_main` (or any sx-defined `#jni_class`) bodied method,
`super.method(args)` lowers to JNI's nonvirtual dispatch against the
parent class resolved via `#extends` (default `android.app.Activity`).
- lower.zig: tracks `current_foreign_class` + `current_foreign_method`
around each `synthesizeJniMainStub` body; pushes the JNIEnv* arg
onto the lexical `#jni_env` stack so omitted-env JNI calls inside
the body see env without a wrapper. New `lowerSuperCall` handles
the `super.method(args)` receiver pattern: derives parent path,
reuses the enclosing method's signature when names match (the
common `super.<override>(args)` case), or looks up the method on
the parent class declared as `#foreign #jni_class`.
- inst.zig: `JniMsgSend` gains `is_nonvirtual: bool` and
`parent_class_path: ?[]const u8` — the dispatch tag + super class
foreign path. Mutually exclusive with `is_static`.
- emit_llvm.zig: new `CallNonvirtual<T>Method` vtable slots + a
fourth dispatch arm. Resolves the parent jclass via
`FindClass(env, parent_path)` (per-call; caching is follow-up),
then `GetMethodID(env, parent_cls, name, sig)`, then
`CallNonvirtual<T>Method(env, obj, parent_cls, mid, args...)`.
Disassembly on the smoke confirms the chain:
`ldr [env+0x30]` (FindClass) → `ldr [env+0x108]` (GetMethodID) →
`ldr [env+0x2d8]` (CallNonvirtualVoidMethod) with `(env, self,
parent_cls, mid, bundle)`.
132 host / 5 cross / zig build test all green. The slice unblocks
Activity lifecycle overrides (onCreate, onResume, onPause) calling
their required `super.<method>(args)` without raw `#jni_call`
boilerplate.
Deletes the entire NativeActivity / native_app_glue / ALooper stack
that the previous Android entry path was built around:
- `examples/99-android-egl-clear.sx` — the demo of the legacy path.
- `library/modules/platform/android.sx` — `AndroidPlatform.init`,
`run_frame_loop`, `sx_android_bootstrap`, `g_android_app`, plus
the ALooper / AInputEvent / ANativeActivity / AConfiguration
foreign decls that fed them. The JNI helpers (`sx_load_javavm_fn`,
`sx_android_get_env`, `sx_query_safe_insets_jni`, the
`ANATIVEACTIVITY_*` offsets) were tied to the ANativeActivity*
delivered to `android_main` — they're stale now that the OS hands
sx code a Java Activity directly via `onCreate(JNIEnv*, jobject)`.
- `library/vendors/sx_android_jni/sx_android_jni.c` — the input-
handler installer (`sx_android_install_input_handler`), which
poked NDK app-pointer field offsets that no longer exist.
`library/modules/platform/android_jni.sx` (the `Activity`/`Window`/
`View`/`WindowInsets` `#jni_class` registry used for safe-insets
dispatch) survives — it's standalone declarative bindings useful from
any `#jni_main` onCreate body. Docstring updated to drop the
"imported from android.sx" framing.
131 host / 4 cross / zig build test all green. End-to-end smoke APK
still produces the expected JNI-mangled symbol +
SxApp-extends-Activity dex.
External consumers (chess) will need to migrate their entry from the
`AndroidPlatform.run_frame_loop` model to the `#jni_main` model
(Java-side Activity drives lifecycle; onSurfaceChanged / Choreographer
drive frames via JNI callbacks). That migration is downstream work.
`checkRequiredEntryPoints` no longer accepts `android_main` as an
Android entry — `#jni_main #jni_class("...")` is now the sole accepted
path. The diagnostic walks the user through declaring an Activity +
Bundle foreign decl. `isExportedEntryName` drops `android_main` and
`ANativeActivity_onCreate` (both were for the legacy NativeActivity
glue path R.2 stopped linking by default).
Migrates the two cross-compile examples that previously carried an
`android_main` trampoline to a minimal `#jni_main #jni_class(...) { }`
stub:
- `examples/ffi-jni-call-02-void.sx` — tests `#jni_call(void)` lowering
- `examples/ffi-objc-call-10-os-gate.sx` — tests `inline if OS` gating
Both stubs are empty (no `onCreate` body) so they exercise the
entry-point check + R.3's JNI-symbol synthesis pass produces no
symbols. 131 host / 4 cross / zig build test all green.
`examples/99-android-egl-clear.sx` still uses `android_main` and the
AndroidPlatform/native_app_glue stack — its Android-target build now
fails the entry-point check. R.5 removes it along with the rest of
the legacy NativeActivity surface.
Loosens lower.zig's `checkRequiredEntryPoints` to accept either a
`#jni_main #jni_class("...")` decl OR the legacy `android_main`
trampoline. The diagnostic now shows both options when neither is
present.
Updates the slice 2 smoke (`examples/ffi-jni-main-01-emit.sx`) to
express the modern shape — drops `android_main`, declares
`Bundle :: #foreign #jni_class("android/os/Bundle")`, and overrides
`onCreate :: (self: *Self, b: *Bundle) { }` inside the #jni_main class.
The emitted Java now correctly declares `void onCreate(Bundle b)` as
@Override + a matching `private native void sx_onCreate(Bundle b)`
delegate, verified via dexdump.
Full retirement of `android_main` (deleting native_app_glue from the
Android link path, dropping `AndroidPlatform.run_frame_loop`, migrating
chess/EGL demo to the Java-driven lifecycle) is multi-slice rework
and stays as follow-up.
Compilation.lowering_jni_main_decls is populated by lowerToIR (iterating
foreign_class_map for is_main && !is_foreign && runtime==jni_class,
deduped by foreign_path); each entry carries the pre-rendered Java source
from jni_java_emit.emitJavaSource.
createApk extended: when the emission list is non-empty, write each
.java under <stage>/java/<pkg>/<Class>.java, javac --release 11 to
<stage>/classes/, d8 --release --lib <android_jar> --output <stage>
to produce <stage>/classes.dex, then zip the .dex into the unaligned
APK at root level. javac discovery: $JAVA_HOME/bin/javac first, then
`which javac`.
Manifest still hardcodes android.app.NativeActivity (slice 3 wires the
user's class name + android:hasCode="true"), so the bundled .dex is
present but unreferenced at runtime. End-to-end verified via dexdump on
the smoke example's APK — Lco/swipelab/sxjnimain/SxApp; extending
NativeActivity shows up in classes.dex. Non-#jni_main APK builds
(99-android-egl-clear.sx) produce the same shape as before.
Cross-compile tuple added for examples/ffi-jni-main-01-emit.sx
(compile-only — APK exercise is manual).
`#jni_call` collapses to a single surface — env is *always* implicit:
either picked up from the lexically-enclosing `#jni_env(env) { ... }`
block's Ref (cheap, register-resident, no TL touch) or from the
runtime's thread-local slot via `sx_jni_env_tl_get()` (one fn call
per dispatch). The explicit-env shape is gone — chess and the
existing tests migrate cleanly by wrapping their helper-fn bodies
in `#jni_env(env) { ... }`.
The TL slot lives outside the user's IR module so the LLVM ORC JIT
can load object files cleanly without `orc_rt` for TLS support:
library/vendors/sx_jni_runtime/sx_jni_env_tl.c:
static _Thread_local void *sx_jni_env_tl_slot;
void *sx_jni_env_tl_get(void) { return sx_jni_env_tl_slot; }
void sx_jni_env_tl_set(void *env) { sx_jni_env_tl_slot = env; }
Linkage:
- sx-the-compiler links the .c file via build.zig so the JIT
process-symbol generator resolves `sx_jni_env_tl_get`/`_set`.
- AOT targets get the same .c file auto-linked via the lowering
pass: when lower touches the TL externs, it sets
`needs_jni_env_tl_runtime`, and `Compilation.lowerToIR` appends a
synthetic `CImportInfo` to `lowering_extra_c_sources` that
`collectCImportSources` merges with user-written ones.
Lowering-side changes:
- `getJniEnvTlFids` lazily declares the two externs (parallel
to `getSelRegisterNameFid`) and flips `needs_jni_env_tl_runtime`.
- `#jni_env(env) { body }` emits save→set→body→restore via three
`call` ops to the externs; the inner body sees env via the
lexical-direct stack.
- `lowerJniCall` resolves env from `jni_env_stack` (top) or the TL
fallback. The explicit-env branch is gone.
- `jni_env_stack_base` tracks per-fn lexical scope so lazy-lowering
a callee doesn't accidentally see the caller's Ref (Refs are only
valid inside one fn's instruction stream).
Test migration (mechanical):
- ffi-jni-call-{01..09}: each helper fn wraps `#jni_call(...)`
bodies in `#jni_env(env) { ... }`. Returning values pass through
the block as an expression — `#jni_env` now also lowers in
expression position.
Verified:
- zig build test + tests/run_examples.sh: 130/130 green.
- tests/cross_compile.sh: 3/3 green.
- Chess APK rebuilt + reinstalled on Pixel. Board renders with
status-bar clearance + info panel intact; no crashes in logcat.
Safe-insets dispatch through `#jni_env` + lexical-direct now
fully exercised end-to-end on real hardware.
A `#jni_call(void)(target, "name", "sig")` inside a helper fn that
isn't lexically inside a `#jni_env` block should fall back to a
thread-local env read populated by the enclosing `#jni_env(env) {
helper(target); }` scope at runtime. Today the lower-side
"jni_env_stack empty" diagnostic gets queued but compilation
continues to emit_llvm, which fails LLVM verification because env
lowers to `Ref.none` (`i64 undef`).
The make-green follow-up:
- Synthesizes a thread-local `@sx_jni_env_tl` global in emit_llvm.
- `#jni_env(env) { body }` emits a `(load TL → saved, store env → TL,
defer store saved → TL)` sequence so the TL tracks the
innermost-scope env and restores correctly on nesting.
- `lowerJniCall`'s omitted-env path falls back to a TL load when
`jni_env_stack` is empty, instead of erroring.
The lexical-direct optimisation from 2.16b stays the fast path —
helpers in the same fn never touch TL. Only cross-fn callees pay
the (cheap) TL load.
Flip the surface semantics for type-introducer directives: bare
`Foo :: #jni_class("path") { ... }` now means "DEFINE a new Java class
at that path" (sx-side provides the implementations). The `#foreign`
prefix modifier flips it back to "REFERENCE an existing class on the
foreign runtime." Matches how `#foreign` already reads in sx for C
function declarations (`printf :: ... #foreign;`).
Foo :: #foreign #jni_class("path/to/Foo") { ... } // reference
Foo :: #jni_class("path/to/Foo") { ... } // define
Foo :: #jni_main #jni_class("path/to/Foo") { ... } // define + main Activity
Compiler-side changes:
- New `hash_jni_main` lexer token (the launchable-Activity marker).
Existing `hash_foreign` is reused; no new modifier token there.
- `ForeignClassDecl` gains `is_foreign: bool` + `is_main: bool`.
`ForeignMethodDecl` gains `body: ?*Node` so defined-class methods
can carry sx-side implementations (foreign-class methods stay
`;`-terminated).
- Parser learns `tryParseForeignClassPrefix` — peek-and-consume the
modifier tokens, then dispatch to the unchanged
`parseForeignClassDecl` with the flags threaded through.
- Sema rejects two illegal combinations: `#foreign + #jni_main`
(can't be both an external reference and the app's main entry),
and bodied methods on `#foreign` decls (foreign methods are
runtime-provided).
- Lower's foreign-class dispatch errors on non-foreign decls with
a pointer to the runtime-synthesis follow-up; defined-class
codegen (Java class emission, RegisterNatives wiring, manifest
entry generation) lands in a separate session.
Migration:
- `library/modules/platform/android_jni.sx`: all four foreign class
decls (`Activity`, `Window`, `View`, `WindowInsets`) gain `#foreign`.
- `examples/ffi-jni-class-{01..08}*.sx`: every test's `#jni_class` /
`#jni_interface` / `#objc_class` / `#objc_protocol` / `#swift_class`
/ `#swift_struct` / `#swift_protocol` usage gains `#foreign`. All
9 files mechanical perl rename; snapshots unchanged.
Verified locally:
- `zig build test` clean.
- `bash tests/run_examples.sh` 129/129.
- `bash tests/cross_compile.sh` 3/3.
- Chess APK rebuilds, reinstalls, launches on Pixel; safe-area
clearance preserved.
`act.getWindow()` on `act: *Activity` (where `Activity ::
#jni_class("android/app/Activity") { getWindow :: ... }`) should
lower to `#jni_call(*void)(act, "getWindow", "()Ljava/lang/Object;")`
(omitted-env form picking up env from the enclosing `#jni_env`
scope via 2.16b's lexical-direct path). Today's sema reports
"unresolved: 'getWindow'" because foreign-class members aren't
yet wired into the method-resolution path.
The make-green follow-up needs:
- sema: register `ForeignClassDecl.members` so method names
resolve on foreign-class receivers (or suppress the unresolved
fallback for them).
- lower: build a `foreign_class_map` in scan pass; new arm in
`lowerCall`'s method-dispatch site emits a synthetic
`FfiIntrinsicCall { kind: jni_call, args: [target, "name",
"(sig)Ret", method_args...] }` with the descriptor derived via
`jni_descriptor.deriveMethod`.
- type system: `*Activity` resolution path so `inferExprType`
on the receiver returns a known type (likely register foreign
classes as synthetic 0-field structs reusing the struct-type
machinery).
Larger session needed — pausing here at the xfail.
`#jni_call(void)(target, "name", "sig")` (3 args before the first
string literal) should work inside an enclosing `#jni_env(env) { ... }`
scope, picking up the env from the block's value directly. Today's
lowering expects 4+ args and errors with "#jni_call requires env,
target, method name, and signature".
The make-green follow-up adds a lowering-side env stack maintained
across the `#jni_env` body walk, and a disambiguation in
`lowerJniCall` that detects "env omitted" via the position of the
first string-literal arg (method name at index 1 → omitted; at index
2 → explicit env).
`#jni_env(synth_env) { ... }` should parse as a block-scoped env
intrinsic, today the lexer doesn't know the directive and the parser
errors at the `#` token in expression position. The make-green
follow-up adds the `hash_jni_env` lexer token, parser arm in
parsePrimary, AST node, and sema acceptance — body runs as a normal
block, env captured for later. TL push/pop semantics + optional env
in `#jni_call` land in 2.16b.
#jni_interface, #objc_class, #objc_protocol, #swift_class,
#swift_struct, #swift_protocol — each with the same body grammar as
#jni_class. Today the lexer doesn't recognise any of these directives
and the parser errors at the first one (`#jni_interface`). The
make-green follow-up adds the six lexer tokens and refactors
`JniClassDecl` into `ForeignClassDecl` with a `runtime` discriminator
so all seven forms share one AST shape and one parser path.
New `hash_jni_method_descriptor` lexer token + LSP keyword
classification. `JniMethodDecl` gains `desc_override: ?[]const u8`.
parseJniClassDecl accepts an optional `#jni_method_descriptor("...")`
clause between the return type and the terminating `;`, stashing the
literal as the override. Auto-derivation in Phase 2.8 will treat
this as the precedence override when present.
The 2.6 xfail commit (0ed4799) used the working name `#desc` in its
test file; this commit renames to `#jni_method_descriptor` for
parallel naming with the rest of the FFI directive set (`#jni_call`,
`#jni_class`, `#jni_env`, ...). Test snapshot flips xfail → green.
125/125 examples green.
`weirdMethod :: (self: *Self) -> s32 #desc("()I");` should parse,
today's 2.5 parser expects `;` immediately after the return type
and errors at the `#desc` token. The make-green follow-up adds a
`hash_desc` lexer token and threads an optional `desc_override`
field through `JniMethodDecl`.
`Point :: #jni_class("...") { x: s32; y: s32; }` should parse,
today's 2.4 body loop sees the identifier `x`, expects `::`, hits
`:` and errors. The make-green follow-up adds a `field` variant to
`JniClassMember` and a parser branch that detects `<ident>:` (vs
`<ident>::`) as the field-decl indicator.
`Window :: #jni_class("...") { #extends View; ... }` should parse,
today's 2.3 parser doesn't recognise `#extends` as a token and the
body loop reports "expected method name". The make-green follow-up
adds `hash_extends`/`hash_implements` lexer tokens, refactors
`JniClassDecl.methods` into a `members` tagged union, and dispatches
in the body loop on the leading token.
`Math :: #jni_class("java/lang/Math") { static abs :: (n: s32) -> s32; }`
should parse, today's 2.2 parser treats `static` as a plain
identifier and errors at the following `abs`. The make-green
follow-up adds a `static` keyword recognition step in the body
loop and an `is_static` flag on `JniMethodDecl`.
Today's parser doesn't recognize #jni_class as a hash directive
after `::`, so it falls through to expression parsing and errors
at the `#` token. Step 2.1 extends parseConstBinding to accept
the directive (opaque on empty body) and re-snapshots this file
to green.
Test-add for static dispatch — `#jni_static_call(s32)(env, cls,
"max", "(II)I", 3, 7)` exercises GetStaticMethodID + CallStaticIntMethod
plus two integer args. Today the lowering bails on `is_static = true`
with `LLVMGetUndef`. IR snapshot captures the placeholder.
The next commit:
- Adds `Jni.GetStaticMethodID` (113), `Jni.CallStaticVoidMethod` (141),
`Jni.CallStaticIntMethod` (129), etc. to the constants struct.
- Wires the static path: skip `GetObjectClass` (`target` IS the
jclass), `NewGlobalRef(target)` to cache it, `GetStaticMethodID`
for the method, then `CallStatic<Type>Method` per return type.
Last return-type variant in the matrix. JNI's jobject is a pointer
(LocalRef) — sx's `*void` maps to LLVM `ptr` directly. CallObjectMethod
is at vtable slot 34. IR snapshot captures today's `ret ptr undef`.
Next commit adds the `.ptr => Jni.CallObjectMethod` arm.
LocalRef lifetime: the returned jobject is a JNI LocalRef bounded by
the native frame. Chains of calls within one frame consume LocalRefs
inline; calls that need to escape the frame should be promoted via
`NewGlobalRef` (already wired in the slot-interning path). Step 1.22
doesn't introduce automatic cleanup — chess use consumes objects
inline, matching the pattern in sx_android_jni.c.
Test-add for the jboolean return. JNI `jboolean` is a single byte (0
or 1); sx's `bool` lowers to LLVM `i1` with byte-coercion at the ABI
boundary. CallBooleanMethod is at vtable slot 37.
IR snapshot captures today's `ret i1 undef`. Next commit adds the
`.bool => Jni.CallBooleanMethod` arm.
Test-add for the jdouble return-type variant — `#jni_call(f64)(env,
target, "getValue", "()D")`. First non-integer return type for JNI.
IR snapshot captures today's `ret double undef` placeholder. The
next commit adds the `.f64 => Jni.CallDoubleMethod` arm.
Test-add for the jlong return-type variant — same shape as 1.18's
jint test but exercising `#jni_call(s64)(env, target,
"currentTimeMillis", "()J")`. Today the non-void switch falls
through to `LLVMGetUndef`; the IR snapshot captures the placeholder.
The next commit adds the `.s64 => Jni.CallLongMethod` arm. The
snapshot will update to show the full dispatch through vtable slot
52, reusing the 1.17 slot interning machinery.
Adds `examples/ffi-jni-call-04-jint-return.sx` exercising
`#jni_call(s32)(env, target, "getCount", "()I")` inside a runtime-
reachable but never-invoked helper (`g_should_call` stays false, so
the dereferences don't fire). Today the emit_llvm switch falls
through to `LLVMGetUndef` for any non-void return — the IR snapshot
captures that placeholder.
The next commit adds the `.s32 => 49` (CallIntMethod) arm. The
snapshot will update to show the full GetObjectClass → GetMethodID →
CallIntMethod sequence (reusing the slot interning landed in 1.17,
since `("getCount", "()I")` is a fresh literal pair).
Adds `examples/ffi-jni-call-03-methodid-sharing.sx` with two
`#jni_call` sites against the same (class, method, sig). Today each
site emits its own `GetObjectClass` + `GetMethodID` + `Call<Type>Method`
sequence (8 vtable indirections total for the two-call test); 1.17
will collapse the two `GetMethodID` calls into a single cached
`jmethodID` static slot populated at module init, mirroring the
`OBJC_SELECTOR_REFERENCES_*` shape that 1.5 introduced for `#objc_call`.
Runtime is a no-op — `unused_jni` is reachable through a
runtime-readable `g_should_call` global that stays false, so the JNI
dereferences never execute. A plain `if false` would get
constant-folded, taking the function definition out of the IR
entirely; the global keeps both the function and its body present
for the IR-snapshot harness.
IR snapshot at `tests/expected/ffi-jni-call-03-methodid-sharing.ir`
locks the pre-caching shape. The next commit (1.17) updates it to the
collapsed shape.
113/113 host tests pass.
Adds `examples/ffi-jni-call-02-void.sx` exercising `#jni_call(void)
(env, target, "name", "sig")` inside an `inline if OS == .android`
arm, plus a new tuple in `tests/cross_compile.sh`. Host run_examples
passes (the inline-if strips the JNI body, leaving "skipped"); the
Android cross-compile FAILs because `lowerFfiIntrinsicCall` still
emits the placeholder diagnostic for any `fic.kind != .objc_call`.
Per the FFI cadence rule this is a test-add (xfail); the next
commit makes the Android cross-compile green by adding the
`.jni_msg_send` opcode and its emit_llvm expansion.
Closes the runtime-verification gap from cluster 1.32. The migrated
`uikit_keyboard_will_change_frame` body uses both shapes but isn't
reached by chess startup (the soft keyboard doesn't open without user
input), so runtime verification was transitive only: `#objc_call(CGRect)`
via the structurally-identical `#objc_call(UIEdgeInsets)` (4×f64 HFA)
in ffi-objc-call-07, and `#objc_call(u64)` via the LLVM-equivalent
`#objc_call(s64)` `hash` test in ffi-objc-call-04.
This example installs two IMPs via `class_addMethod`:
- `rect_imp` returns a CGRect of {10.5, 20.5, 30.5, 40.5} through the
32-byte HFA path (v0..v3 on AAPCS64).
- `u64_imp` returns `0x7FEDCBA987654321` through the i64 path.
`#objc_call(CGRect)` and `#objc_call(u64)` dispatch through them and
the values are printed for snapshot lockdown.
Reused the parser quirk noted in the checkpoint and in 0.1 — integer
literals ≥ 2^63 are rejected even when the receiving type is u64, so
the test value keeps the high bit clear.
111/111 host tests pass.
`collectCaptures` in `src/ir/lower.zig` was the closure free-variable
analyzer that decides which names from a closure body need to be
boxed into the env struct at lambda-build time. Its switch on AST
node kind enumerated every other shape (`.call`, `.if_expr`,
`.match_expr`, `.for_expr`, etc.) but no arm for `.ffi_intrinsic_call`,
so the trailing `else => {}` quietly dropped its `args[]` and
`return_type` walks. Names referenced inside `#objc_call(T)(recv,
"sel:", ...)` from a closure body never made it into the captures
list, so when lowering bound the closure scope from env, those names
came back as "unresolved".
The fix adds the missing arm — walk `return_type` and every `args[i]`
the same way `.call` walks `callee` + `args`.
Companion changes:
- `examples/issue-0038.sx` → `examples/103-ffi-closure-capture.sx`
(out of the open-issue namespace; comment header tightened to
describe the feature, not the historical bug).
- `examples/ffi-objc-call-09-in-construct.sx` drops the
`g_hasher_recv` module-global workaround that was added for this
bug — the closure now captures `recv` from `make_hasher`'s arg
list normally.
Uncomments the second passthrough case in `examples/issue-0038.sx`
that captures `recv` from the enclosing function into a closure body
that uses it inside `#objc_call(s64)(recv, "hash")`. Current behavior
is a hard error from the name-resolution pass:
examples/issue-0038.sx:28:48: error: unresolved: 'recv'
Snapshot locks the failure in (exit 1 + that error message) so the
next commit can flip it to passing without ambiguity. Per the FFI
cadence rule this is a test-add (xfail); the make-green follow-up
adds the missing recursion arm in `lower.zig`'s `collectCaptures` for
`.ffi_intrinsic_call` nodes.
Closes the runtime-verification gap from cluster 1.28: chess startup
doesn't reach the keyboard `becomeFirstResponder` / `resignFirstResponder`
path, so `#objc_call(bool)` was only compile-verified. This example
installs two BOOL-returning IMPs via `class_addMethod` (type encoding
"B@:") and dispatches both through `#objc_call(bool)`. Also exercises
the nil-receiver guarantee (libobjc returns a zero slot, which decodes
as false).
This is a test-add commit (per the FFI cadence rule): it locks in
current behavior without changing any lowering. Lowering shape is
identical to `#objc_call(u8)` at the ABI layer; this test makes the
source-level type explicit and gives `git bisect` a target if a
future emit_llvm change inadvertently breaks single-byte returns.
110/110 host tests pass.
109/109 regression tests pass; chess Android + iOS-sim still
build clean.
Root cause: sx's `xx <ptr>` cast targeting an integer type
(common pattern: `xx u64 = xx @some_global`) lowered to a no-op
because `coerceToType` had branches for int↔float and same-kind
widen/narrow, but nothing for pointer↔integer. The cast left the
value as a pointer Ref, and `emitInst`'s `.ret` arm tried to
coerce a `ptr` value to an `i64` slot — coerceArg had no
ptr↔int branch either, fell through to undef.
Why it worked in main but failed in helpers: an
`alloca u64`+`store ptr @g, alloca`+`load i64, alloca` sequence
preserves the address bits as raw memory, so the
"store-then-load through an alloca" workaround happened to do
the right thing without a real cast. A `ret i64 <ptr>` has no
such intermediate slot and triggers an LLVM type mismatch.
Fix layered into two existing IR opcodes:
lower.zig (coerceToType):
new branch — when src and dst types are ptr↔int, emit a
`bitcast` IR opcode with the right from/to. Mirrors how
int↔float emits `.int_to_float` / `.float_to_int`.
emit_llvm.zig (.bitcast arm):
dispatch ptr→int to `LLVMBuildPtrToInt` (+ trunc/zext if the
target int width != 64), int→ptr to `LLVMBuildIntToPtr`. The
"real bitcast" path stays for same-kind type punning.
Modern LLVM's BuildBitCast rejects ptr↔int directly, hence
the dispatch.
The fix also closes a quiet behavior gap that affected non-`#foreign`
globals (any `xx @<global>` from a helper fn). Surfaced while
investigating issue-0037; verified independently with a
non-`#foreign` sx-side global of type `s64`.
File mechanics: issue-0037 promoted to a focused feature example
per CLAUDE.md's resolution flow:
examples/issue-0037.sx -> examples/102-foreign-global-from-helper.sx
tests/expected/issue-0037.{txt,exit} -> tests/expected/102-foreign-global-from-helper.{txt,exit}
ffi-objc-call-03 + ffi-objc-call-06 IR snapshots updated to
reflect the ptr→int store-via-ptrtoint shape that's now correct
at the LLVM-IR level (same bits in memory, but properly typed).
109/109 host tests pass; tests/cross_compile.sh's first real tuple
(`android | examples/ffi-objc-call-10-os-gate.sx`) compiles
through `sx build --target android` without finding any
`@objc_msgSend` / `@sel_registerName` symbols in the output —
the `inline if OS == .ios { #objc_call(...) }` arm is stripped
at sx compile time before emit_llvm runs, so the Android
toolchain (Bionic + libGLESv3 / NDK linker) doesn't see the
Obj-C runtime references that would otherwise be undefined.
Host (macOS): the example prints "host stripped both" — the iOS
arm is stripped (we're not iOS) AND the Android arm is stripped
(we're not Android), confirming `inline if OS == { case }`
symmetric strip-and-render works around `#objc_call` sites.
The example carries a 3-line `android_main` trampoline so the
NDK linker's `-u ANativeActivity_onCreate` / entry-point
discovery is satisfied — pattern shared with chess + the other
android examples.
108/108 regression tests pass (+ffi-objc-call-09-in-construct,
+issue-0038 from the prior commit).
One trivial Obj-C call (`[obj hash]` returning NSUInteger) routed
through four sx surface constructs:
1. struct method body Probe.fetch
2. protocol impl method body impl Hashable for Probe
3. closure value body make_hasher
4. generic function body hash_through(recv: $T)
No new ABI shapes touched — pins that the `objc_msg_send` lowering
emits identical call shapes regardless of enclosing scope. Each
case validates the result `h_N == h_1` after threading `recv`
appropriately for each context.
The closure path reaches `recv` via a module-level global rather
than capturing the surrounding parameter — issue-0038 (prior
commit) documents the closure free-variable analyzer missing the
`FfiIntrinsicCall` node, with a clean workaround pinned.
Surfaced while writing the Phase 1.11 in-construct test. The
closure free-variable analyzer doesn't recursively visit the
`ffi_intrinsic_call` AST node introduced in Phase 1.1, so any
identifier used inside `#objc_call` / `#jni_call` /
`#jni_static_call` from a closure body trips:
error: unresolved: '<name>'
The same identifier captured from the same scope into a plain
expression resolves fine — so the bug is localized to whatever
recursive arm-walk powers the capture analysis.
Likely fix: add an `ffi_intrinsic_call => { ... }` arm wherever
the `.call =>` arm visits `callee` + `args`. Candidate files:
- src/sema.zig (capture / scope tracking)
- src/ir/lower.zig (closure body lowering / `lowerLambda`)
Both should be checked.
Workaround in the meantime: reach the captured value via a
module-level global from inside the closure body. See the
`g_hasher_recv` pattern in
examples/ffi-objc-call-09-in-construct.sx for an applied
instance.
106/106 regression tests pass (+ffi-objc-call-08-multi-keyword).
`#objc_call(s32)(instance, "combine:and:", 7, 42)` round-trips
end-to-end via class_addMethod-registered IMP that does
`a * 100 + b` → 742. Pins three things:
1. The two-keyword selector "combine:and:" parses, mangles, and
interns under the symbol `@OBJC_SELECTOR_REFERENCES_combine_and_`
(every `:` → `_` — matches clang).
2. Multi-arg call lowering correctly puts arg0 / arg1 in the right
slots after recv / sel.
3. The IMP-side sx fn signature `(self, _cmd, a: s32, b: s32)`
with `callconv(.c)` interops with the Obj-C runtime's typical
IMP shape, and the runtime forwards the keyword args to the
right physical positions.
No codegen change — Phase 1.6's variadic-args branch in the
`objc_msg_send` lowering already handled this; this test just
locks in the surface.
105/105 regression tests pass (+ffi-objc-call-07-fp-hfa-return).
Same round-trip pattern as 1.8 — register an Obj-C class at
runtime with class_addMethod, IMP returns specific non-zero values,
#objc_call reads them back — but for an all-double 32 B HFA
instead of a 24 B int aggregate.
Locks in the f32-vs-f64 landmine that bit us when we first
wrote safeAreaInsets in uikit.sx: the homogeneous-float-aggregate
ABI routes 1..4 f32 or f64 fields through v0..v3 (AAPCS64) /
xmm0..xmm3 (SysV AMD64) WITHOUT integer coercion. As long as the
LLVM call-site function type carries the precise struct (which
our `objc_msg_send` arm does), the backend lowers it correctly.
This is the smaller cousin of 1.8 — 1.8 needed an emit_llvm code
change to make the sret transform work; 1.9 needs no codegen
change because HFAs of any size up to v0..v3 stay register-resident.
The test just pins that path with a real, value-bearing IMP so a
future ABI-rule shake-up has a regression net.
104/104 regression tests pass. The Triple round-trip
(triple_imp writes {11, 22, 33} on the IMP side → #objc_call(Triple)
reads them back) is the test of record.
emit_llvm.zig changes:
1. `objc_msg_send` arm — when `needsByval(ret_ty)` (same predicate
the plain-foreign-call path uses), apply the sret transform:
- ret type collapses to void
- prepend a `ptr` param at index 0 (call site provides an
alloca slot)
- mirror `sret(<RetType>)` on the call site so the AArch64 x8
/ SysV-AMD64 hidden-ptr ABI lowers correctly
- load the result from the slot post-call
The IR shape now matches clang exactly:
call void @objc_msgSend(ptr sret({...}) %slot, ptr %recv, ptr %sel)
2. `.ret` arm — the body-side counterpart for sx fns whose declared
return type is sret-shaped (sx-defined IMPs registered via
`class_addMethod` produce these). When the current function's
`needsByval(func.ret)` predicate holds, store the IR ret value
through the prepended sret slot (param 0) and emit `ret void`.
Previously the unconditional coerceArg path turned the struct
value into `undef` and emitted `ret void undef` — illegal LLVM.
Test mechanics: registers `SxTripleProbe : NSObject` at runtime via
`objc_allocateClassPair` + `class_addMethod`, IMP returns
Triple{11, 22, 33}. `#objc_call(Triple)(instance, "tripleValue")`
gets them back, round-trip pinned in the .txt snapshot and the
IR-shape snapshot.
103/103 regression tests pass (+ffi-objc-call-06-sret-return).
The runtime output is misleadingly clean — `[nil tripleValue]`
zeros all three fields because libobjc's nil-stub clears the
return registers. But the IR snapshot reveals the actual ABI
mismatch:
%objc.msg = call { i64, i64, i64 } @objc_msgSend(ptr null, ptr %load)
A live receiver returning a non-zero `Triple` would surface
garbage in the third field — the AArch64 backend lowers
{ i64, i64, i64 } returns to x0/x1 pair + a third register that
the runtime's sret-shaped stub doesn't populate.
Next commit (1.8b): emit_llvm's `objc_msg_send` arm gains the
same sret transform we did for plain `#foreign` calls in Phase
0.3 — ret type collapses to void, prepend a ptr sret param,
alloca the result slot at the call site, mirror the
`sret(<T>)` attribute on the call, load result from the slot
post-call. IR snapshot will flip to:
%slot = alloca <Triple>
call void @objc_msgSend(ptr sret(<Triple>) %slot, ptr null, ptr %load)
%objc.msg = load <Triple>, ptr %slot
103/103 regression tests pass (+ffi-objc-call-05-struct-returns).
Three return shapes all round-trip cleanly with the existing Phase
1.6 `objc_msg_send` lowering — no codegen change needed because
emit_llvm.zig hands the IR struct type straight to LLVMBuildCall2
and the AArch64 / SysV AMD64 backends already know how to lower:
NSPoint — 16 B HFA (2×f64) → v0, v1 (AAPCS64) / xmm0, xmm1 (SysV)
NSRange — 16 B 2×u64 → x0, x1 register pair via [2 x i64]
NSRect — 32 B HFA (4×f64) → v0..v3 (AAPCS64) / xmm0..xmm3 (SysV)
Verified against the Obj-C runtime's `[nil structMethod]`-returns-
zero contract — no real-object setup needed, but the wider ABI
path runs exactly as it would for live receivers (the registers
the runtime stub uses come back through the same lowering).
>16 B non-HFA aggregates (e.g. {3×s64}) trip a sret cliff and
land in Phase 1.8. Verified locally that they return garbage in
the trailing field today — register pair / quad won't carry the
extra storage, and emit_llvm's `objc_msg_send` arm doesn't apply
the sret transform yet.
102/102 regression tests pass; chess Android + iOS-sim still build
clean. `ffi-objc-call-04-primitive-returns` flips from xfail to
passing with both nil-recv and real-recv flavors of *void / s64
returns exercised.
Key change: a new `objc_msg_send` IR opcode bundles (recv, sel,
extra args) and carries the return type via the `Inst.ty` field.
emit_llvm.zig builds a per-call-site LLVM function type from the
argument Refs' IR types (recv/sel as ptr; extra args through
abiCoerceParamType) and dispatches with LLVMBuildCall2. One
declared `@objc_msgSend` symbol is reused across every return
type — opaque pointers make the function value type-erased, so
each call site picks its own ABI.
before: one (recv, sel) -> ptr LLVM declaration, hard-coded
per call site; only void return wired in 1.3.
after: same declaration, each call site provides a fresh
LLVMBuildCall2 fn-type → s64 / *void / bool / f64
returns all dispatch correctly without separate FuncIds.
Selector init mechanism: stayed with the @llvm.global_ctors
constructor. Investigated clang's
`__DATA,__objc_selrefs` + `externally_initialized` shape — works
for fully-linked binaries (dyld substitutes the SEL at load
time) but **LLVM ORC JIT** (the engine behind `sx run`) doesn't
process Mach-O Obj-C metadata sections, so the slot keeps its
initial value (the method-name string pointer) and dispatch
crashes with "<null selector>". The portable choice: keep the
constructor AND inject a direct call to it at `main`'s entry —
idempotent under dyld (sel_registerName returns the same SEL on
re-registration), required for ORC JIT.
Files touched:
src/ir/inst.zig | new ObjcMsgSend struct + opcode
src/ir/lower.zig | drop the void-only restriction; emit the
new opcode; remove the orphaned
getObjcMsgSendFid path (objc_msgSend
declaration moved to emit_llvm)
src/ir/emit_llvm.zig | objc_msg_send arm (per-call-site
LLVMBuildCall2); lazy `@objc_msgSend`
declaration via getObjcMsgSendValue;
emitObjcSelectorInit refactored to inject
the ctor call at main's entry
src/ir/{print,interp}.zig | switch arms for the new opcode
`ffi-objc-call-03-selector-sharing.ir` snapshot updates to
reflect the new shape (the `call ... @objc_msgSend` call sites
no longer mention a typed wrapper).
102/102 regression tests pass (+ffi-objc-call-04-primitive-returns
with xfail snapshot capturing today's diagnostic).
Pinned scenario: `[NSObject class]` — `#objc_call(*void)(null, "class")`.
Should return a non-null Class pointer once the lowering supports
non-void returns. Today the Phase 1.3 restriction trips with:
#objc_call: only `void` return + (recv, selector) is lowered today;
non-void / arg-bearing arities land in later phase-1 steps
The next commit (1.6b) introduces an `objc_msg_send` IR opcode that
bundles (recv, sel, args, ret_ty) and emit_llvm builds a per-call-
site LLVM function type, sharing one declared `@objc_msgSend`
symbol across return-type variants. Five primitive returns
(*void / bool / s32 / s64 / f64) get folded in across 1.6b–c.
run_examples.sh now supports an optional `tests/expected/<name>.ir`
sibling to `.txt`/`.exit`. When present, the runner also captures
`sx ir <file>` output, normalizes target-/host-specific noise
(module ID, target triple/datalayout, attribute groups, LLVM's
auto-suffixed %temp numbering), and diffs against the snapshot.
`--update` regenerates it alongside the runtime output.
Catches lowering changes that don't affect what the program prints
— exactly the shape Phase 1.5's selector interning will produce
(same runtime output, very different IR).
First snapshot: `ffi-objc-call-03-selector-sharing.ir`. Today the
test emits four `call ptr @sel_registerName(ptr @str.N)` lines for
its four call sites; after 1.5 we expect two static
`@OBJC_SELECTOR_REFERENCES_<sel>` globals + loads at each call
site. The diff between the two snapshots will be the visible
artifact of the optimization.
101/101 regression tests pass (+ffi-objc-call-03-selector-sharing).
Test exercises four call sites — three sharing "init" and one
"release" — to pin the multi-site / multi-selector lowering before
1.5 changes how SEL lookups are cached.
Runtime behavior: identical before and after 1.5 (all call sites
hit nil receivers; libobjc returns 0 for void). The improvement is
visible only in the emitted IR — today:
$ ./zig-out/bin/sx ir examples/ffi-objc-call-03-selector-sharing.sx \\
| grep -c "call ptr @sel_registerName"
4
After 1.5 (planned): 2 — one `sel_registerName` per unique selector
string, materialized into a static `OBJC_SELECTOR_REFERENCES_<sel>`
global at module init, then loaded at each call site. Matches the
shape clang produces for `@selector(...)`. Worth re-running the
above grep after 1.5 lands as a manual sanity check.
The IR-shape snapshot harness (auto-diff of `sx ir` output) is
deferred; for now we verify by eye.
100/100 regression tests pass (+ffi-objc-call-02-void-return xfail
snapshot).
The intrinsic with no `inline if false` guard reaches sema/codegen
and trips an "unresolved: 'unknown_expr'" — the FfiIntrinsicCall
AST node from Phase 1.1 has no lowering rules in lower.zig /
emit_llvm.zig yet.
nil receiver was chosen so the test doesn't need a real Obj-C
object graph: the runtime guarantees `[nil msg]` is a no-op with
zero result for void returns. macOS-gated via `inline if OS == .macos`
so the runner stays portable.
Next commit: emit_llvm.zig produces the per-call-site
%sel = call ptr @sel_registerName(ptr "init.0")
call void @objc_msgSend(ptr null, ptr %sel)
lowering. Snapshot flips to "ok". Selector interning (one shared
global per unique selector string) lands as a separate step (1.5).
