Further narrowed it down to the "--no-lazy" runtime flag:
Working:
std::string flags = "--expose_gc --jitless";
V8::SetFlagsFromString(flags.c_str(), flags.size());
NOT working:
std::string flags = "--expose_gc --jitless --no-lazy";
V8::SetFlagsFromString(flags.c_str(), flags.size());
Any idea why would the --no-lazy flag affect function names in stack
frames? I need the --no-lazy flag because I noticed that it speeds up code
cached scripts (v8::ScriptCompiler::CreateCodeCache API).
On Tuesday, November 19, 2019 at 9:30:07 AM UTC+2, Darin Dimitrov wrote:
>
> Simon, thanks for the clarification. Does this mean that there's no
> reliable way to get this information with the current V8 public API?
>
> I went ahead and wrote a simple example illustrating the issue and tried
> running it on different platforms. And the results were surprising.
>
> Here's the sample code I wrote:
>
> Isolate::CreateParams create_params;
> create_params.array_buffer_allocator = &allocator_;
> Isolate* isolate = Isolate::New(create_params);
> Isolate::Scope isolate_scope(isolate);
> HandleScope handle_scope(isolate);
> Local<ObjectTemplate> globalTemplate = ObjectTemplate::New(isolate);
> Local<Context> context = Context::New(isolate, nullptr,
> globalTemplate);
> context->Enter();
>
> Local<v8::Function> captureFunc = v8::Function::New(context, [](const
> FunctionCallbackInfo<Value>& info) {
> Isolate* isolate = info.GetIsolate();
> Local<Context> context = isolate->GetCurrentContext();
> Local<Object> exception =
> Exception::Error(v8::String::NewFromUtf8(isolate, "",
> NewStringType::kNormal, 0).ToLocalChecked()).As<Object>();
> Local<Value> stackProperty = exception->Get(context,
> v8::String::NewFromUtf8(isolate, "stack", NewStringType::kNormal,
> 5).ToLocalChecked()).ToLocalChecked();
> v8::String::Utf8Value stackStr(isolate, stackProperty);
> printf("%s\n\n", *stackStr);
>
> Local<StackTrace> stack =
> v8::StackTrace::CurrentStackTrace(isolate, 10,
> v8::StackTrace::StackTraceOptions::kDetailed);
>
> for (int i = 0; i < stack->GetFrameCount(); i++) {
> Local<v8::StackFrame> frame = stack->GetFrame(isolate, i);
> Local<v8::String> funcName = frame->GetFunctionName();
> v8::String::Utf8Value str(isolate, funcName);
> const char* name = *str;
> printf("%s (%d:%d)\n", name, frame->GetLineNumber(),
> frame->GetColumn());
> }
> }).ToLocalChecked();
> assert(context->Global()->Set(context,
> v8::String::NewFromUtf8(isolate, "captureTrace", NewStringType::kNormal,
> 12).ToLocalChecked(), captureFunc).FromMaybe(false));
>
> std::string src =
> "(function test() {"
> " const viewModel = new Object();"
> " viewModel.onTap = () => { captureTrace(); };"
> " viewModel.onTap.apply(viewModel, []);"
> "})();";
> Local<Script> script = Script::Compile(context,
> v8::String::NewFromUtf8(isolate, src.c_str(), NewStringType::kNormal,
> (int)src.length()).ToLocalChecked()).ToLocalChecked();
> Local<Value> result;
> assert(script->Run(context).ToLocal(&result));
>
> On V8 8.0.0 (
> https://chromium.googlesource.com/v8/v8.git/+/50031fae736fac7b2b309369df492bfd0edd7553),
>
> running in --jitless on an arm64 iOS device, I get the following output:
>
> Error
> at Object.onTap (<anonymous>:1:84)
> at test (<anonymous>:1:122)
> at <anonymous>:1:145
>
> (null) (1:84)
> test (1:122)
> (null) (1:145)
>
> and on V8 7.7.299.11 (
> https://chromium.googlesource.com/v8/v8.git/+/027689dbfcb2a9bbc8ceec4db2631c558e879633),
>
> running on an arm64 Android device, I get the following output:
>
> Error
> at Object.viewModel.onTap (<anonymous>:1:84)
> at test (<anonymous>:1:122)
> at <anonymous>:1:145
>
> viewModel.onTap (1:84)
> test (1:122)
> (null) (1:145)
>
> Do you know what might explain the discrepancy between those two platforms?
>
> On Tuesday, November 19, 2019 at 7:58:10 AM UTC+2, Simon Zünd wrote:
>>
>> This is a bug in our API implementation. Or better, there is a mismatch
>> between API and what we do internally. "Method names" and "class names" are
>> stored separately from function names. This can be seen here, when we
>> collect all the information for a single frame:
>> https://cs.chromium.org/chromium/src/v8/src/heap/factory.cc?rcl=e1eb815647f334a8cf970439343a8febfa9f6d11&l=3732
>>
>> There is really no reason why this information shouldn't be made
>> available in the API. The only tricky thing is to come up with a better
>> API, as some of the getters change meaning depending whether its a JS or
>> WASM frame and I am not really happy with the current solution.
>>
>> On Mon, Nov 18, 2019 at 4:40 PM Darin Dimitrov <darin....@gmail.com>
>> wrote:
>>
>>> I am embedding V8 and trying to capture the current javascript
>>> stacktrace using the "v8::StackTrace::CurrentStackTrace":
>>>
>>> Isolate* isolate = info.GetIsolate();
>>>
>>> Local<StackTrace> stack = v8::StackTrace::CurrentStackTrace(isolate, 10,
>>> v8::StackTrace::StackTraceOptions::kDetailed);
>>>
>>> *for* (*int* i = 0; i < stack->GetFrameCount(); i++) {
>>>
>>> Local<v8::StackFrame> frame = stack->GetFrame(isolate, i);
>>>
>>> Local<v8::String> funcName = frame->GetFunctionName();
>>>
>>> v8::String::Utf8Value str(isolate, funcName);
>>>
>>> *const* *char** name = *str;
>>>
>>> printf("%s (%d:%d)\n", name, frame->GetLineNumber(), frame->
>>> GetColumn());
>>>
>>> }
>>>
>>> I have placed this code inside a custom function that I have registered
>>> to the global scope.
>>>
>>> Before calling my custom function, I tried logging the stack property of
>>> an Error object (by using : console.log(new Error().stack)) and I obtained
>>> the following output:
>>>
>>> Error
>>> at bar (file:///app/bundle.js:266:29)
>>> at test (file:///app/bundle.js:267:15)
>>> at Observable.onTap (file:///app/bundle.js:268:11)
>>> at Button.notify (file:///app/vendor.js:3620:32)
>>> at Button._emit (file:///app/vendor.js:3640:18)
>>> at TapHandlerImpl.tap (file:///app/vendor.js:15540:19)
>>>
>>> With the "v8::StackTrace::CurrentStackTrace" method I get:
>>>
>>> bar (266:29)
>>> test (267:15)
>>> (null) (268:11)
>>> (null) (3620:32)
>>> (null) (3640:18)
>>> (null) (15540:19)
>>>
>>> I am getting empty function names from some frames, while all the other
>>> information is present (script name, line, column numbers, ...).
>>>
>>> Do you know what might be the reason for getting empty function names?
>>>
>>> --
>>> --
>>> v8-users mailing list
>>> v8-u...@googlegroups.com
>>> http://groups.google.com/group/v8-users
>>> ---
>>> You received this message because you are subscribed to the Google
>>> Groups "v8-users" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to v8-u...@googlegroups.com.
>>> To view this discussion on the web visit
>>> https://groups.google.com/d/msgid/v8-users/64d9e4ca-b271-4d83-aace-753c28e193f4%40googlegroups.com
>>>
>>> <https://groups.google.com/d/msgid/v8-users/64d9e4ca-b271-4d83-aace-753c28e193f4%40googlegroups.com?utm_medium=email&utm_source=footer>
>>> .
>>>
>>
>
--
--
v8-users mailing list
v8-users@googlegroups.com
http://groups.google.com/group/v8-users
---
You received this message because you are subscribed to the Google Groups
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to v8-users+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/v8-users/358d6517-c279-459b-a675-9ad4d712638d%40googlegroups.com.
