Segfault with experimental modules and mocha

[mattgodbolt]

• Version: v12.4.0
• Platform: Linux godbolt.aq.tc 5.1.8-300.fc30.x86_64 deps: update openssl to 1.0.1j #1 SMP Sun Jun 9 17:09:32 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux
• Subsystem: Looks like v8

I'm running node with --experimental-modules --es-module-specifier-resolution=node and against a codebase made up of .mjs files. I have a small shim to make this play nicely with mocha, and the presence of a particular import statement crashes v8. Depending on which files I exclude/include it differs, which makes this hard to debug. Commenting in or out a particular import triggers it, however.

The code is unfortunately proprietary, and I've as yet not been able to repro with a small test case.

I built a debug node and the stack trace is:

Thread 1 "node" received signal SIGSEGV, Segmentation fault.
0x0000000000dcdd6a in v8::Local<v8::Function>::New (isolate=0x39073f0, that=...)
    at ../deps/v8/include/v8.h:9691
9691	  return New(isolate, that.val_);
(gdb) bt
#0  0x0000000000dcdd6a in v8::Local<v8::Function>::New (isolate=0x39073f0, that=...)
    at ../deps/v8/include/v8.h:9691
#1  0x0000000000dccd95 in v8::PersistentBase<v8::Function>::Get (this=0x10, isolate=0x39073f0)
    at ../deps/v8/include/v8.h:482
#2  0x0000000000dc9147 in node::loader::ImportModuleDynamically (context=..., referrer=..., 
    specifier=...) at ../src/module_wrap.cc:979
#3  0x000000000176c66a in v8::internal::Isolate::RunHostImportModuleDynamicallyCallback (
    this=this@entry=0x39073f0, referrer=..., specifier=specifier@entry=...)
    at ../deps/v8/include/v8.h:315
#4  0x0000000001a83605 in v8::internal::__RT_impl_Runtime_DynamicImportCall (args=..., 
    isolate=0x39073f0) at ../deps/v8/src/runtime/runtime-module.cc:28
#5  0x0000000001a85677 in v8::internal::Runtime_DynamicImportCall (args_length=2, 
    args_object=0x7fffffff9d78, isolate=0x39073f0)
    at ../deps/v8/src/runtime/runtime-module.cc:15
#6  0x0000000002653ce4 in Builtins_CEntry_Return1_DontSaveFPRegs_ArgvInRegister_NoBuiltinExit
    () at ../deps/v8/src/builtins/builtins-async-iterator-gen.cc:338
#7  0x00000000027c7cd2 in Builtins_CallRuntimeHandler ()
    at ../deps/v8/src/interpreter/interpreter-generator.cc:1733
#8  0x00007fffffff9d28 in ?? ()
#9  0x00007fffffff9d28 in ?? ()
#10 0x00007fffffff9db8 in ?? ()
#11 0x000001ea6231ec59 in ?? ()
#12 0x0000000000000018 in ?? ()
#13 0x00007fffffff9db8 in ?? ()
#14 0x000000000238f535 in Builtins_InterpreterEntryTrampoline ()
    at ../deps/v8/src/builtins/builtins-string-gen.cc:2081
#15 0x0000126a363804d1 in ?? ()
#16 0x0000126a363804d1 in ?? ()
#17 0x000001ea6231ec59 in ?? ()
#18 0x000028df8921f789 in ?? ()
#19 0x000027fcbb682579 in ?? ()
#20 0x00001fb5bbf5a4b1 in ?? ()
#21 0x00001fb5bbf5a3e1 in ?? ()
#22 0x000000eb00000000 in ?? ()
#23 0x000001ea6231f139 in ?? ()
#24 0x000028df8921f789 in ?? ()
#25 0x000027fcbb682579 in ?? ()
#26 0x00007fffffff9df8 in ?? ()
#27 0x0000000002454d4c in Builtins_AsyncFunctionAwaitResolveClosure ()
    at ../deps/v8/src/builtins/builtins-async-function-gen.cc:224
Backtrace stopped: previous frame inner to this frame (corrupt stack?)

I repro this with (to remove some of the noise from running mocha directly, though this has no bearing on the bug itself.

$ gdb --args ~/dev/node-v12.4.0/out/Debug/node --experimental-modules --es-module-specifier-resolution=node ./node_modules/.bin/_mocha ./test/test.esm.js

The test.esm.js is:

(async () => {
 /* usually this is a bunch of globbing and auto importing, but this is a minimal repro, and this is a subset of files */
    await import('/home/mgodbolt/dev/coral/web/octopus/test/datascope/datascope-service-test.mjs');
    await import('/home/mgodbolt/dev/coral/web/octopus/test/datascope/nodes-test.mjs');
    await import('/home/mgodbolt/dev/coral/web/octopus/test/graph-test.mjs');
    run();
})();

The offending line that causes the issue is in datascope-service-test.mjs, which I've commented out save the single line:

import {PendingExtraction} from "../../src/lib/datascope/datascope-service";

and this triggers the issue.

Please let me know how I can best help with more information. Thanks!

[mscdex]

Duplicate of #27492.

[mattgodbolt]

Thanks! I did search but somehow didn't find that issue!