Add optional primitive verification of JS library dependencies (#23855)

This is not enabled by default, and perhaps never will be. The changes
in this PR are the resulting the running whole test suite with
`EMCC_CHECK_DEPS=1` set in the envionment.

Author: sbc100

src/jsifier.mjs

@@ -45,6 +45,10 @@ const addedLibraryItems = {};
 
 const extraLibraryFuncs = [];
 
+// Experimental feature to check for invalid __deps entries.
+// See `EMCC_CHECK_DEPS` in in the environment to try it out.
+const CHECK_DEPS = process.env.EMCC_CHECK_DEPS;
+
 // Some JS-implemented library functions are proxied to be called on the main
 // browser thread, if the Emscripten runtime is executing in a Web Worker.
 // Each such proxied function is identified via an ordinal number (this is not
@@ -184,6 +188,60 @@ function preJS() {
   return result;
 }
 
+// Certain library functions have specific indirect dependencies.  See the
+// comments alongside eaach of these.
+const checkDependenciesSkip = new Set([
+  '_mmap_js',
+  '_emscripten_throw_longjmp',
+  '_emscripten_receive_on_main_thread_js',
+  'emscripten_start_fetch',
+  'emscripten_start_wasm_audio_worklet_thread_async',
+]);
+
+const checkDependenciesIgnore = new Set([
+  // These are added in bulk to whole library files are so are not precise
+  '$PThread',
+  '$WebGPU',
+  '$SDL',
+  '$GLUT',
+  '$GLEW',
+  '$Browser',
+  '$AL',
+  '$GL',
+  '$IDBStore',
+  // These are added purely for their side effects
+  '$polyfillWaitAsync',
+  '$GLImmediateSetup',
+  '$emscriptenGetAudioObject',
+  // These get conservatively injected via i53ConversionDeps
+  '$bigintToI53Checked',
+  '$convertI32PairToI53Checked',
+  'setTempRet0',
+]);
+
+/**
+ * Hacky attempt to find unused `__deps` entries.  This is not enabled by default
+ * but can be enabled by setting CHECK_DEPS above.
+ * TODO: Use a more precise method such as tokenising using acorn.
+ */
+function checkDependencies(symbol, snippet, deps, postset) {
+  if (checkDependenciesSkip.has(symbol)) {
+    return;
+  }
+  for (const dep of deps) {
+    if (typeof dep === 'function') {
+      continue;
+    }
+    if (checkDependenciesIgnore.has(dep)) {
+      continue;
+    }
+    const mangled = mangleCSymbolName(dep);
+    if (!snippet.includes(mangled) && (!postset || !postset.includes(mangled))) {
+      error(`${symbol}: unused dependency: ${dep}`);
+    }
+  }
+}
+
 function addImplicitDeps(snippet, deps) {
   // There are some common dependencies that we inject automatically by
   // conservatively scanning the input functions for their usage.
@@ -576,6 +634,18 @@ function(${args}) {
       let isFunction = false;
       let aliasTarget;
 
+      const postsetId = symbol + '__postset';
+      const postset = LibraryManager.library[postsetId];
+      if (postset) {
+        // A postset is either code to run right now, or some text we should emit.
+        // If it's code, it may return some text to emit as well.
+        const postsetString = typeof postset == 'function' ? postset() : postset;
+        if (postsetString && !addedLibraryItems[postsetId]) {
+          addedLibraryItems[postsetId] = true;
+          postSets.push(postsetString + ';');
+        }
+      }
+
       if (typeof snippet == 'string') {
         if (snippet[0] != '=') {
           if (LibraryManager.library[snippet]) {
@@ -594,19 +664,8 @@ function(${args}) {
         isFunction = true;
         snippet = processLibraryFunction(snippet, symbol, mangled, deps, isStub);
         addImplicitDeps(snippet, deps);
-      }
-
-      const postsetId = symbol + '__postset';
-      let postset = LibraryManager.library[postsetId];
-      if (postset) {
-        // A postset is either code to run right now, or some text we should emit.
-        // If it's code, it may return some text to emit as well.
-        if (typeof postset == 'function') {
-          postset = postset();
-        }
-        if (postset && !addedLibraryItems[postsetId]) {
-          addedLibraryItems[postsetId] = true;
-          postSets.push(postset + ';');
+        if (CHECK_DEPS && !isUserSymbol) {
+          checkDependencies(symbol, snippet, deps, postset?.toString());
         }
       }
 

src/lib/libaddfunction.js

@@ -84,7 +84,9 @@ addToLibrary({
     }
   },
   // Wraps a JS function as a wasm function with a given signature.
+#if !WASM2JS
   $convertJsFunctionToWasm__deps: ['$uleb128Encode', '$sigToWasmTypes', '$generateFuncType'],
+#endif
   $convertJsFunctionToWasm: (func, sig) => {
 #if WASM2JS
     // return func;
@@ -192,8 +194,12 @@ addToLibrary({
   $addFunction__docs: '/** @param {string=} sig */',
   $addFunction__deps: ['$convertJsFunctionToWasm', '$getFunctionAddress',
                        '$functionsInTableMap', '$getEmptyTableSlot',
-                       '$getWasmTableEntry', '$setWasmTableEntry',
-                       '$wasmTable'],
+                       '$setWasmTableEntry',
+#if ASSERTIONS >= 2
+                       '$getWasmTableEntry', '$wasmTable',
+#endif
+  ],
+
   $addFunction: (func, sig) => {
   #if ASSERTIONS
     assert(typeof func != 'undefined');

src/lib/libautodebug.js

@@ -18,7 +18,7 @@ addToLibrary({
   $get_i64__deps: ['setTempRet0'],
   $get_i64: (loc, index, low, high) => {
     dbg('get_i64 ' + [loc, index, low, high]);
-    setTempRet0(high);
+    _setTempRet0(high);
     return low;
   },
   $get_f32: (loc, index, value) => {
@@ -52,7 +52,7 @@ addToLibrary({
   $set_i64__deps: ['setTempRet0'],
   $set_i64: (loc, index, low, high) => {
     dbg('set_i64 ' + [loc, index, low, high]);
-    setTempRet0(high);
+    _setTempRet0(high);
     return low;
   },
   $set_f32: (loc, index, value) => {
@@ -90,7 +90,7 @@ addToLibrary({
   $load_val_i64__deps: ['setTempRet0'],
   $load_val_i64: (loc, low, high) => {
     dbg('load_val_i64 ' + [loc, low, high]);
-    setTempRet0(high);
+    _setTempRet0(high);
     return low;
   },
   $load_val_f32: (loc, value) => {
@@ -112,7 +112,7 @@ addToLibrary({
   $store_val_i64__deps: ['setTempRet0'],
   $store_val_i64: (loc, low, high) => {
     dbg('store_val_i64 ' + [loc, low, high]);
-    setTempRet0(high);
+    _setTempRet0(high);
     return low;
   },
   $store_val_f32: (loc, value) => {

src/lib/libccall.js

@@ -130,7 +130,11 @@ addToLibrary({
    * @param {Array=} argTypes
    * @param {Object=} opts
    */`,
-  $cwrap__deps: ['$getCFunc', '$ccall'],
+  $cwrap__deps: [ '$ccall',
+#if !ASSERTIONS
+    '$getCFunc',
+#endif
+  ],
   $cwrap: (ident, returnType, argTypes, opts) => {
 #if !ASSERTIONS
     // When the function takes numbers and returns a number, we can just return

src/lib/libcore.js

@@ -202,15 +202,15 @@ addToLibrary({
   },
 
   emscripten_resize_heap__deps: [
-    '$getHeapMax',
-    '$alignMemory',
-#if ASSERTIONS == 2
-    'emscripten_get_now',
-#endif
 #if ABORTING_MALLOC
     '$abortOnCannotGrowMemory',
 #endif
 #if ALLOW_MEMORY_GROWTH
+#if ASSERTIONS == 2
+    'emscripten_get_now',
+#endif
+    '$getHeapMax',
+    '$alignMemory',
     '$growMemory',
 #endif
   ],
@@ -791,7 +791,7 @@ addToLibrary({
     return str;
   },
 
-  $readSockaddr__deps: ['$Sockets', '$inetNtop4', '$inetNtop6', 'ntohs'],
+  $readSockaddr__deps: ['$inetNtop4', '$inetNtop6', 'ntohs'],
   $readSockaddr: (sa, salen) => {
     // family / port offsets are common to both sockaddr_in and sockaddr_in6
     var family = {{{ makeGetValue('sa', C_STRUCTS.sockaddr_in.sin_family, 'i16') }}};
@@ -825,7 +825,7 @@ addToLibrary({
     return { family: family, addr: addr, port: port };
   },
   $writeSockaddr__docs: '/** @param {number=} addrlen */',
-  $writeSockaddr__deps: ['$Sockets', '$inetPton4', '$inetPton6', '$zeroMemory', 'htons'],
+  $writeSockaddr__deps: ['$inetPton4', '$inetPton6', '$zeroMemory', 'htons'],
   $writeSockaddr: (sa, family, addr, port, addrlen) => {
     switch (family) {
       case {{{ cDefs.AF_INET }}}:
@@ -914,7 +914,7 @@ addToLibrary({
     return inetPton4(DNS.lookup_name(nameString));
   },
 
-  getaddrinfo__deps: ['$Sockets', '$DNS', '$inetPton4', '$inetNtop4', '$inetPton6', '$inetNtop6', '$writeSockaddr', 'malloc', 'htonl'],
+  getaddrinfo__deps: ['$DNS', '$inetPton4', '$inetNtop4', '$inetPton6', '$inetNtop6', '$writeSockaddr', 'malloc', 'htonl'],
   getaddrinfo__proxy: 'sync',
   getaddrinfo: (node, service, hint, out) => {
     // Note getaddrinfo currently only returns a single addrinfo with ai_next defaulting to NULL. When NULL
@@ -1085,7 +1085,7 @@ addToLibrary({
     return 0;
   },
 
-  getnameinfo__deps: ['$Sockets', '$DNS', '$readSockaddr', '$stringToUTF8'],
+  getnameinfo__deps: ['$DNS', '$readSockaddr', '$stringToUTF8'],
   getnameinfo: (sa, salen, node, nodelen, serv, servlen, flags) => {
     var info = readSockaddr(sa, salen);
     if (info.errno) {
@@ -1251,8 +1251,7 @@ addToLibrary({
   // Timers always fire on the main thread, either directly from JS (here) or
   // or when the main thread is busy waiting calling _emscripten_yield.
   _setitimer_js__proxy: 'sync',
-  _setitimer_js__deps: ['$timers', '$callUserCallback',
-                        '_emscripten_timeout', 'emscripten_get_now'],
+  _setitimer_js__deps: ['$timers', '$callUserCallback', '_emscripten_timeout', 'emscripten_get_now'],
   _setitimer_js: (which, timeout_ms) => {
 #if RUNTIME_DEBUG
     dbg(`setitimer_js ${which} timeout=${timeout_ms}`);
@@ -1506,12 +1505,7 @@ addToLibrary({
 #endif
 
   $readEmAsmArgsArray: [],
-  $readEmAsmArgs__deps: [
-    '$readEmAsmArgsArray',
-#if MEMORY64
-    '$readI53FromI64',
-#endif
-  ],
+  $readEmAsmArgs__deps: ['$readEmAsmArgsArray'],
   $readEmAsmArgs: (sigPtr, buf) => {
 #if ASSERTIONS
     // Nobody should have mutated _readEmAsmArgsArray underneath us to be something else than an array.
@@ -1778,7 +1772,11 @@ addToLibrary({
 #endif
     return f(ptr, ...args);
   },
-  $dynCall__deps: ['$dynCallLegacy', '$getWasmTableEntry'],
+#if DYNCALLS
+  $dynCall__deps: ['$dynCallLegacy'],
+#else
+  $dynCall__deps: ['$getWasmTableEntry'],
+#endif
 #endif
 
   // Used in library code to get JS function from wasm function pointer.
@@ -2101,7 +2099,7 @@ addToLibrary({
 #if PTHREADS
     '_emscripten_thread_exit',
 #endif
-#if RUNTIME_DEBUG
+#if RUNTIME_DEBUG >= 2
     '$runtimeKeepaliveCounter',
 #endif
   ],
@@ -2159,7 +2157,9 @@ addToLibrary({
 
   // Allocate memory for an mmap operation. This allocates space of the right
   // page-aligned size, and clears the allocated space.
+#if hasExportedSymbol('emscripten_builtin_memalign')
   $mmapAlloc__deps: ['$zeroMemory', '$alignMemory'],
+#endif
   $mmapAlloc: (size) => {
 #if hasExportedSymbol('emscripten_builtin_memalign')
     size = alignMemory(size, {{{ WASM_PAGE_SIZE }}});
@@ -2416,6 +2416,7 @@ function wrapSyscallFunction(x, library, isWasi) {
   // has disabled the filesystem or we have proven some other way that this will
   // not be called in practice, and do not need that code.
   if (!SYSCALLS_REQUIRE_FILESYSTEM && t.includes('FS.')) {
+    library[x + '__deps'] = [];
     t = modifyJSFunction(t, (args, body) => {
       return `(${args}) => {\n` +
              (ASSERTIONS ? "abort('it should not be possible to operate on streams when !SYSCALLS_REQUIRE_FILESYSTEM');\n" : '') +

src/lib/libdylink.js

@@ -203,7 +203,7 @@ var LibraryDylink = {
   },
 
   $updateGOT__internal: true,
-  $updateGOT__deps: ['$GOT', '$isInternalSym', '$addFunction', '$getFunctionAddress'],
+  $updateGOT__deps: ['$GOT', '$isInternalSym', '$addFunction'],
   $updateGOT: (exports, replace) => {
 #if DYLINK_DEBUG
     dbg("updateGOT: adding " + Object.keys(exports).length + " symbols");
@@ -945,7 +945,7 @@ var LibraryDylink = {
   // flags.global and flags.nodelete are handled every time a load request is made.
   // Once a library becomes "global" or "nodelete", it cannot be removed or unloaded.
   $loadDynamicLibrary__deps: ['$LDSO', '$loadWebAssemblyModule',
-                              '$isInternalSym', '$mergeLibSymbols', '$newDSO',
+                              '$mergeLibSymbols', '$newDSO',
                               '$asyncLoad',
 #if FILESYSTEM
                               '$preloadedWasm',
@@ -1119,7 +1119,7 @@ var LibraryDylink = {
   },
 
   // void* dlopen(const char* filename, int flags);
-  $dlopenInternal__deps: ['$ENV', '$dlSetError', '$PATH'],
+  $dlopenInternal__deps: ['$dlSetError', '$PATH'],
   $dlopenInternal: (handle, jsflags) => {
     // void *dlopen(const char *file, int mode);
     // http://pubs.opengroup.org/onlinepubs/009695399/functions/dlopen.html