Thank you very much for your advice, really appreciate it.
在 2020年3月31日星期二 UTC+8下午5:36:59，Jakob Kummerow写道：
>
> This is expected. FunctionCallbackInfo instances are rarely constructed
> via C++ `new`; instead they are usually created manually on the stack. So
> adding a field to the class definition is going to be a lot of work, and
> that field's type can't be a std::string -- it would have to be something
> where you have precise control over the object layout.
>
> I would suggest that you explore alternative approaches.
>
>
> On Fri, Mar 27, 2020 at 6:21 AM L TY <n0b0...@gmail.com <javascript:>>
> wrote:
>
>> Hi everyone,
>>
>> For some research reason, I need to log all property access to Blink in
>> v8. Since all the property load information is passed to Blink through
>> class `FunctionCallbackInfo`, I tried to add a pointer field in this class
>> to hold the property name (represented as std::string) like follow:
>>
>> V8_INLINE FunctionCallbackInfo(internal::Address* implicit_args,
>> internal::Address* values, int length);
>> internal::Address* implicit_args_;
>> internal::Address* values_;
>> int length_;
>> std::string* property_info_; // The field I add.
>>
>> }
>>
>> The `property_info_` field is initialized in the constructor function
>> through c++ `new`.
>> I also added a const setter function for the `property_info_` field. The
>> setter function like follow:
>>
>> void FunctionCallbackInfo<T>::SetPropertyInfo(::std::string s) const {
>> *property_info_ = s;
>> }
>>
>> In Blink, I modified the template file to set this field every time an
>> attribute in Blink is visited.
>> In code snips I write by myself, it works well. However, when I try to
>> run the code on some real-world websites, I found that the value of the
>> field property_info_ is set to a strange value 0x38, so the pointer to the
>> string is overlapped by some other value.
>> I tried to mark the property_info_ field as private and even moved it to
>> class `ReturnValue<T>`, the overlapping case still happens:
>>
>> template<typename T>
>> class ReturnValue {
>> public:
>> template <class S> V8_INLINE ReturnValue(const ReturnValue<S>& that)
>> : value_(that.value_) {
>> TYPE_CHECK(T, S);
>> }
>> // Local setters
>> template <typename S>
>> V8_INLINE void Set(const Global<S>& handle);
>> template <typename S>
>> V8_INLINE void Set(const TracedGlobal<S>& handle);
>> template <typename S>
>> V8_INLINE void Set(const Local<S> handle);
>> // Fast primitive setters
>> V8_INLINE void Set(bool value);
>> V8_INLINE void Set(double i);
>> V8_INLINE void Set(int32_t i);
>> V8_INLINE void Set(uint32_t i);
>> // Fast JS primitive setters
>> V8_INLINE void SetNull();
>> V8_INLINE void SetUndefined();
>> V8_INLINE void SetEmptyString();
>> // Convenience getter for Isolate
>> V8_INLINE Isolate* GetIsolate() const;
>>
>> // Pointer setter: Uncompilable to prevent inadvertent misuse.
>> template <typename S>
>> V8_INLINE void Set(S* whatever);
>>
>> // Getter. Creates a new Local<> so it comes with a certain performance
>> // hit. If the ReturnValue was not yet set, this will return the
>> undefined
>> // value.
>> V8_INLINE Local<Value> Get() const;
>> // use for taint map by n0b0dy
>> V8_INLINE internal::Address GetValueAddr() const;
>>
>> V8_INLINE void SetProperty(std::string) const; // The setter function I
>> add
>>
>> private:
>> template<class F> friend class ReturnValue;
>> template<class F> friend class FunctionCallbackInfo;
>> template<class F> friend class PropertyCallbackInfo;
>> template <class F, class G, class H>
>> friend class PersistentValueMapBase;
>> V8_INLINE void SetInternal(internal::Address value) { *value_ = value; }
>> V8_INLINE internal::Address GetDefaultValue();
>> V8_INLINE explicit ReturnValue(internal::Address* slot);
>> internal::Address* value_;
>> internal::Address* property_info_; // the field I add
>> };
>>
>>
>>
>> I read the in
>> https://source.chromium.org/chromium/chromium/src/+/master:v8/src/builtins/x64/builtins-x64.cc;drc=34ad93361f96ad875bf350ced27e15b7aa126113;bpv=1;bpt=1;l=3317
>>
>> and found that v8 will arrange the stack frame with its own call
>> convention, But I didn't find why the pointer value is still overlapped in
>> class ReturnValue.
>>
>> Does anyone have ideas about what should I do?
>>
>> Thanks a lot!
>>
>> --
>> --
>> v8-dev mailing list
>> v8-...@googlegroups.com <javascript:>
>> http://groups.google.com/group/v8-dev
>> ---
>> You received this message because you are subscribed to the Google Groups
>> "v8-dev" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to v8-...@googlegroups.com <javascript:>.
>> To view this discussion on the web visit
>> https://groups.google.com/d/msgid/v8-dev/96c9afe2-5c55-40fc-86b6-9f34de988809%40googlegroups.com
>>
>> <https://groups.google.com/d/msgid/v8-dev/96c9afe2-5c55-40fc-86b6-9f34de988809%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>
--
--
v8-dev mailing list
v8-dev@googlegroups.com
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups
"v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to v8-dev+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/v8-dev/3bde27cd-2190-47a3-89bd-57e94b3b339a%40googlegroups.com.
