ASYNCIFY_DEBUG causes stack overflow for empty program

[JensMunkHansen]

Create an empty C++ module and create an ES6 module.

Compile options for target Dummy: -fexceptions;-pthread;-matomics;-mbulk-memory;-msimd128;-O0;-g3
Compile definitions for target Dummy: COMPILE_DEFINITIONS-NOTFOUND
Include directories for target Dummy: INCLUDE_DIRECTORIES-NOTFOUND
Link libraries for target Dummy: Threads::Threads
Link options for target Dummy: -sASSERTIONS=2;-sERROR_ON_UNDEFINED_SYMBOLS=1;-sDISABLE_EXCEPTION_CATCHING=0;-sALLOW_BLOCKING_ON_MAIN_THREAD=0;-lembind;-sNO_EXIT_RUNTIME=1;-sMODULARIZE=1;-sEXPORT_ES6=1;-sINCLUDE_FULL_LIBRARY;-sALLOW_TABLE_GROWTH=1;-sEXPORT_NAME=loadDummyTestModule;-sINITIAL_MEMORY=1GB;-sALLOW_MEMORY_GROWTH=0;-sENVIRONMENT=node,worker;-pthread;-flto;--enable-bulk-memory;-sUSE_PTHREADS=1;-sSTACK_SIZE=1048576;-sPTHREAD_POOL_SIZE=2;-sPTHREAD_POOL_SIZE_STRICT=16;-sSHARED_MEMORY=1;-sWASM=1;-sASYNCIFY_STACK_SIZE=81920;-sASYNCIFY=1;-sASYNCIFY_DEBUG=2;-sEXPORTED_RUNTIME_METHODS=['ENV','FS','addFunction','removeFunction','wasmTable','ccall','cwrap','stringToNewUTF8','spawnThread'];-sEXPORTED_FUNCTIONS=['_free','_malloc'];-sASYNCIFY_IMPORTS=[];-O0;-g3
-- Configuring done (0.3s)
-- Generating done (0.0s)
-- Build files have been written to: /home/jmh/git/vtkEmbinding/build
[ 0%] Built target Dummy_copy_run_Dummy.js
[100%] Linking CXX executable Dummy.js

error: library_async.js: failure to execute js library "library_async.js":
error: library_async.js: use -sVERBOSE to see more details
Internal compiler error in src/compiler.mjs!
Please create a bug report at https://github.com/emscripten-core/emscripten/issues/
with a log of the build and the input files used to run. Exception message: "library_async.preprocessed.js:73
dbg(ASYNCIFY: ${' '.repeat(Asyncify.exportCallStack.length} try ${x});
^^^^^^

SyntaxError: missing ) after argument list
at new Script (node:vm:117:7)
at createScript (node:vm:269:10)
at Module.runInContext (node:vm:300:10)
at runInMacroContext (file:///home/jmh/github/emsdk/upstream/emscripten/src/utility.mjs:333:13)
at Object.load (file:///home/jmh/github/emsdk/upstream/emscripten/src/modules.mjs:248:9)
at Module.runJSify (file:///home/jmh/github/emsdk/upstream/emscripten/src/jsifier.mjs:169:18)
at file:///home/jmh/github/emsdk/upstream/emscripten/src/compiler.mjs:89:11
em++: error: '/home/jmh/github/emsdk/node/20.18.0_64bit/bin/node /home/jmh/github/emsdk/upstream/emscripten/src/compiler.mjs /tmp/tmpj8etiry9.json --symbols-only' failed (returned 1)

I discovered another issue. The generated code..

  debugInit() {
        function pthreadLogPrefix() {
          var t = 0;
          if (runtimeInitialized && typeof _pthread_self != 'undefined'
          ) {
            t = _pthread_self();
          }
          return `w:${workerID},t:${ptrToString(t)}: `;
        }
  
        // Prefix all err()/dbg() messages with the calling thread ID.
        var origDbg = dbg;
        dbg = (...args) => origDbg(pthreadLogPrefix() + args.join(' '));
      },

Causes a stack overflow during the early initialization. I patched the generated code and removed dbg and reinserted after the module was fully initialize. The issue only occurs if you use -sASYNCIFY_DEBUG.

I use the latest release 3.1.74

Thanks in advance

[sbc100]

Woops!

Looks like this only effect -sASYNCIFY_DEBUG=2 so -sASYNCIFY_DEBUG should still work until we can land a fix.

[JensMunkHansen]

I had the same infinite loop with ASYNCIFY=1, an empty program

…

On Wed, Jan 8, 2025, 03:13 Sam Clegg ***@***.***> wrote: Woops! Looks like this only effect -sASYNCIFY_DEBUG=2 so -sASYNCIFY_DEBUG should still work until we can land a fix. — Reply to this email directly, view it on GitHub <#23333 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABQJKYXIQDNIBSETDG3RUDT2JSCVBAVCNFSM6AAAAABUY4EI3GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNZWGU3DOOBUGE> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[sbc100]

I see, there is a second issue. The syntax error is only present with -sASYNCIFY_DEBUG=2. That first issue is fixed in #23334