Include heap views only when needed

src/jsifier.mjs

@@ -247,12 +247,35 @@ function addImplicitDeps(snippet, deps) {
     'runtimeKeepalivePush',
     'runtimeKeepalivePop',
     'UTF8ToString',
+    // TODO: Consider removing getValue and setValue if they are rarely used implicitly.
+    'getValue',
+    'setValue',
   ];
   for (const dep of autoDeps) {
     if (snippet.includes(dep + '(')) {
       deps.push('$' + dep);
     }
   }
+  // If the snippet contains eval(), it may dynamically evaluate code loaded from memory at runtime
+  // (for example, in emscripten_run_script where the snippet is eval(UTF8ToString(ptr))).
+  // Because static string matching cannot inspect what strings are stored in memory or evaluated
+  // at runtime, we must conservatively include all heap views whenever a snippet uses eval().
+  if (snippet.includes('eval(')) {
+    deps.push('$HEAP8', '$HEAPU8', '$HEAP16', '$HEAPU16', '$HEAP32', '$HEAPU32', '$HEAPF32', '$HEAPF64');
+    if (WASM_BIGINT || MEMORY64) {
+      deps.push('$HEAP64', '$HEAPU64');
+    }
+  }
+  const heapDeps = [
+    'HEAP8', 'HEAP16', 'HEAPU8', 'HEAPU16',
+    'HEAP32', 'HEAPU32', 'HEAPF32', 'HEAPF64',
+    'HEAP64', 'HEAPU64',
+  ];
+  for (const heap of heapDeps) {
+    if (snippet.includes(heap)) {
+      deps.push('$' + heap);
+    }
+  }
 }
 
 function sigToArgs(sig) {
@@ -408,6 +431,14 @@ export async function runJSify(outputFile, symbolsOnly) {
 
   const symbolsNeeded = DEFAULT_LIBRARY_FUNCS_TO_INCLUDE;
   symbolsNeeded.push(...extraLibraryFuncs);
+
+  for (const fileName of [...PRE_JS_FILES, ...POST_JS_FILES]) {
+    const content = readFile(fileName);
+    addImplicitDeps(content, symbolsNeeded);
+  }
+  for (const snippet of EM_JS_SNIPPETS) {
+    addImplicitDeps(snippet, symbolsNeeded);
+  }
   for (const sym of EXPORTED_RUNTIME_METHODS) {
     if ('$' + sym in LibraryManager.library) {
       symbolsNeeded.push('$' + sym);

src/runtime_common.js

@@ -99,13 +99,22 @@ var runtimeExited = false;
       shouldExport = true;
     }
     return shouldExport;
-  }
+  };
   const maybeExportHeap = (x) => {
     if (shouldExportHeap(x) && MODULARIZE != 'instance') {
       return `Module['${x}'] = `;
     }
     return '';
   };
+  const isHeapNeeded = (x) => {
+    return shouldExportHeap(x) || addedLibraryItems['$' + x];
+  };
+  const updateHeap = (x, type) => {
+    if (isHeapNeeded(x)) {
+      return `${maybeExportHeap(x)}${x} = new ${type}(b);`;
+    }
+    return '';
+  };
 }}}
 
 
@@ -161,17 +170,17 @@ function updateMemoryViews() {
 #endif
   var b = wasmMemory.buffer;
 #endif
-  {{{ maybeExportHeap('HEAP8')   }}}HEAP8 = new Int8Array(b);
-  {{{ maybeExportHeap('HEAP16')  }}}HEAP16 = new Int16Array(b);
-  {{{ maybeExportHeap('HEAPU8')  }}}HEAPU8 = new Uint8Array(b);
-  {{{ maybeExportHeap('HEAPU16') }}}HEAPU16 = new Uint16Array(b);
-  {{{ maybeExportHeap('HEAP32')  }}}HEAP32 = new Int32Array(b);
-  {{{ maybeExportHeap('HEAPU32') }}}HEAPU32 = new Uint32Array(b);
-  {{{ maybeExportHeap('HEAPF32') }}}HEAPF32 = new Float32Array(b);
-  {{{ maybeExportHeap('HEAPF64') }}}HEAPF64 = new Float64Array(b);
+  {{{ updateHeap('HEAP8',   'Int8Array')    }}}
+  {{{ updateHeap('HEAP16',  'Int16Array')   }}}
+  {{{ updateHeap('HEAPU8',  'Uint8Array')   }}}
+  {{{ updateHeap('HEAPU16', 'Uint16Array')  }}}
+  {{{ updateHeap('HEAP32',  'Int32Array')   }}}
+  {{{ updateHeap('HEAPU32', 'Uint32Array')  }}}
+  {{{ updateHeap('HEAPF32', 'Float32Array') }}}
+  {{{ updateHeap('HEAPF64', 'Float64Array') }}}
 #if WASM_BIGINT
-  {{{ maybeExportHeap('HEAP64')  }}}HEAP64 = new BigInt64Array(b);
-  {{{ maybeExportHeap('HEAPU64') }}}HEAPU64 = new BigUint64Array(b);
+  {{{ updateHeap('HEAP64',  'BigInt64Array') }}}
+  {{{ updateHeap('HEAPU64', 'BigUint64Array') }}}
 #endif
 #if SUPPORT_BIG_ENDIAN
   {{{ maybeExportHeap('HEAP_DATA_VIEW') }}} HEAP_DATA_VIEW = new DataView(b);

src/settings_internal.js

@@ -295,5 +295,11 @@ var LOAD_SOURCE_MAP = false;
 
 var ALIASES = [];
 
+// Internal setting for passing EM_JS and EM_ASM function code snippets to JSifier so that
+// implicit heap dependencies (e.g. HEAP8, HEAP32) can be scanned and included automatically
+// without needing to search for them in Python.
+// This could be removed if/when we require explicit heap dependencies (via EM_JS_DEPS).
+var EM_JS_SNIPPETS = [];
+
 // List of public setting names (Used by RETAIN_COMPILER_SETTINGS)
 var PUBLIC_SETTINGS = [];

test/codesize/audio_worklet_wasm.expected.js

@@ -1,4 +1,4 @@
-var m = globalThis.Module || typeof Module != "undefined" ? Module : {}, r = !!globalThis.WorkerGlobalScope, t = !!globalThis.AudioWorkletGlobalScope, u = globalThis.name == "em-ww" || t, v, K, E, G, J, x, W, P, F, D, C, X, A, Z, H;
+var m = globalThis.Module || typeof Module != "undefined" ? Module : {}, r = !!globalThis.WorkerGlobalScope, t = !!globalThis.AudioWorkletGlobalScope, u = globalThis.name == "em-ww" || t, v, x, J, K, G, E, W, P, F, D, C, X, A, Z, H;
 
 function w(a) {
     v = a;

test/codesize/hello_wasm_worker_wasm.expected.js

@@ -1,80 +1,77 @@
-var c = Module, d = !!globalThis.WorkerGlobalScope, e = globalThis.name == "em-ww", f, g, C, q, D, m, E, v;
+var c = Module, d = !!globalThis.WorkerGlobalScope, e = globalThis.name == "em-ww", f, g, B, p, C, l, D, u;
 
 e && (onmessage = a => {
     onmessage = null;
     f = a = a.data;
     g = a.o;
-    h();
     c ||= {};
     c.wasm = a.m;
-    k();
+    h();
     a.m = a.o = 0;
 });
 
-function h() {}
-
 e || (g = new WebAssembly.Memory({
     initial: 256,
     maximum: 256,
     shared: !0
-}), h());
+}));
 
-var l = [], n = a => {
+var k = [], m = a => {
     a = a.data;
     var b = a._wsc;
-    b && m.get(b)(...a.x);
-}, p = a => {
-    l.push(a);
-}, r = () => {
-    q(0, !d, !e, d && 1);
-}, u = {}, w = (a, b, z) => {
-    var t = u[a] = new Worker(c.js, {
+    b && l.get(b)(...a.x);
+}, n = a => {
+    k.push(a);
+}, q = () => {
+    p(0, !d, !e, d && 1);
+}, t = {}, v = (a, b, y) => {
+    var r = t[a] = new Worker(c.js, {
         name: "em-ww"
     });
-    t.postMessage({
+    r.postMessage({
         A: a,
-        m: v,
+        m: u,
         o: g,
         u: b,
-        v: z
+        v: y
     });
-    t.onmessage = n;
+    r.onmessage = m;
     return !0;
-}, x = () => performance.now(), y = () => !1, A = (a, b) => {
-    u[a].postMessage({
+}, w = () => performance.now(), x = () => !1, z = (a, b) => {
+    t[a].postMessage({
         _wsc: b,
         x: []
     });
 };
 
-e && (u[0] = globalThis, addEventListener("message", p));
+e && (t[0] = globalThis, addEventListener("message", n));
 
-function B() {
+function A() {
     console.log("Hello from wasm worker!");
 }
 
-function k() {
-    E = {
-        d: r,
-        c: w,
-        b: x,
-        e: y,
-        f: A,
-        g: B,
+function h() {
+    D = {
+        d: q,
+        c: v,
+        b: w,
+        e: x,
+        f: z,
+        g: A,
         a: g
     };
     WebAssembly.instantiate(c.wasm, {
-        a: E
+        a: D
     }).then((a => {
         var b = (a.instance || a).exports;
-        v = a.module || c.wasm;
-        C = b.i;
-        q = b.k;
-        D = b.l;
-        m = b.j;
-        e ? (D(f.A, f.u, f.v), removeEventListener("message", p), l = l.forEach(n), addEventListener("message", n)) : b.h();
-        e || C();
+        u = a.module || c.wasm;
+        B = b.i;
+        p = b.k;
+        C = b.l;
+        l = b.j;
+        e ? (C(f.A, f.u, f.v), removeEventListener("message", n), k = k.forEach(m), addEventListener("message", m)) : b.h();
+        e || B();
     }));
 }
 
-e || k();
+e || h();

test/codesize/hello_world_wasm.expected.js

@@ -1,21 +1,21 @@
-var d = Module, e, f = new TextDecoder, g, h;
+var c = Module, d = new TextDecoder, e, f, h;
 
-WebAssembly.instantiate(d.wasm, {
+WebAssembly.instantiate(c.wasm, {
     a: {
         a: a => {
-            var c = console, k = c.log;
+            var g = console, k = g.log;
             if (a) {
                 for (var b = a, l = e, m = b + NaN; l[b] && !(b >= m); ) ++b;
-                a = f.decode(e.subarray(a, b));
+                a = d.decode(e.subarray(a, b));
             } else a = "";
-            k.call(c, a);
+            k.call(g, a);
         }
     }
 }).then((a => {
     a = a.instance.exports;
-    g = a.d;
+    f = a.d;
     h = a.b;
     e = new Uint8Array(h.buffer);
     a.c();
-    g();
+    f();
 }));

test/codesize/hello_world_wasm2js.expected.js

@@ -1,4 +1,4 @@
-var d = Module, g, h = new TextDecoder, k, l;
+var d = Module, g = new TextDecoder, h, k, l;
 
 function e(a) {
     this.exports = function(q) {
@@ -48,8 +48,8 @@ function e(a) {
         a: a => {
             var q = console, u = q.log;
             if (a) {
-                for (var n = a, c = g, f = n + void 0; c[n] && !(n >= f); ) ++n;
-                a = h.decode(g.subarray(a, n));
+                for (var n = a, c = h, f = n + void 0; c[n] && !(n >= f); ) ++n;
+                a = g.decode(h.subarray(a, n));
             } else a = "";
             u.call(q, a);
         }
@@ -58,7 +58,7 @@ function e(a) {
     a = a.instance.exports;
     k = a.d;
     l = a.b;
-    g = new Uint8Array(l.buffer);
+    h = new Uint8Array(l.buffer);
     a.c();
     k();
 }));

test/codesize/test_codesize_cxx_ctors1.json

@@ -1,10 +1,10 @@
 {
-  "a.out.js": 19183,
-  "a.out.js.gz": 7971,
+  "a.out.js": 19083,
+  "a.out.js.gz": 7947,
   "a.out.nodebug.wasm": 132603,
   "a.out.nodebug.wasm.gz": 49939,
-  "total": 151786,
-  "total_gz": 57910,
+  "total": 151686,
+  "total_gz": 57886,
   "sent": [
     "__cxa_throw",
     "_abort_js",

test/codesize/test_codesize_cxx_ctors2.json

@@ -1,10 +1,10 @@
 {
-  "a.out.js": 19160,
-  "a.out.js.gz": 7957,
+  "a.out.js": 19060,
+  "a.out.js.gz": 7931,
   "a.out.nodebug.wasm": 132029,
   "a.out.nodebug.wasm.gz": 49599,
-  "total": 151189,
-  "total_gz": 57556,
+  "total": 151089,
+  "total_gz": 57530,
   "sent": [
     "__cxa_throw",
     "_abort_js",

test/codesize/test_codesize_cxx_except.json

@@ -1,10 +1,10 @@
 {
-  "a.out.js": 23158,
-  "a.out.js.gz": 8955,
+  "a.out.js": 23059,
+  "a.out.js.gz": 8939,
   "a.out.nodebug.wasm": 172523,
   "a.out.nodebug.wasm.gz": 57480,
-  "total": 195681,
-  "total_gz": 66435,
+  "total": 195582,
+  "total_gz": 66419,
   "sent": [
     "__cxa_begin_catch",
     "__cxa_end_catch",

test/codesize/test_codesize_cxx_except_wasm.json

@@ -1,10 +1,10 @@
 {
-  "a.out.js": 18982,
-  "a.out.js.gz": 7888,
+  "a.out.js": 18882,
+  "a.out.js.gz": 7863,
   "a.out.nodebug.wasm": 147928,
   "a.out.nodebug.wasm.gz": 55349,
-  "total": 166910,
-  "total_gz": 63237,
+  "total": 166810,
+  "total_gz": 63212,
   "sent": [
     "_abort_js",
     "_tzset_js",

test/codesize/test_codesize_cxx_except_wasm_legacy.json

@@ -1,10 +1,10 @@
 {
-  "a.out.js": 19060,
-  "a.out.js.gz": 7913,
+  "a.out.js": 18960,
+  "a.out.js.gz": 7888,
   "a.out.nodebug.wasm": 145734,
   "a.out.nodebug.wasm.gz": 54976,
-  "total": 164794,
-  "total_gz": 62889,
+  "total": 164694,
+  "total_gz": 62864,
   "sent": [
     "_abort_js",
     "_tzset_js",

test/codesize/test_codesize_cxx_lto.json

@@ -1,10 +1,10 @@
 {
-  "a.out.js": 18526,
-  "a.out.js.gz": 7665,
+  "a.out.js": 18426,
+  "a.out.js.gz": 7640,
   "a.out.nodebug.wasm": 102090,
   "a.out.nodebug.wasm.gz": 39548,
-  "total": 120616,
-  "total_gz": 47213,
+  "total": 120516,
+  "total_gz": 47188,
   "sent": [
     "a (emscripten_resize_heap)",
     "b (_setitimer_js)",