Web Inspector: NMI provide data for mixing with tcmalloc heap dumps

[Ilya Tikhonovsky]

Early we used tcmalloc heap dumps for check what classes have to be instrumented first.
We instrumented a lot of classes and it became difficult to use heap dumps for this task.
We need a way to see the classes that aren't instrumented yet or their objects were not reached during snapshot.
We could use heap dump but need to mix tcmalloc data with the data about already counted objects.

[Ilya Tikhonovsky]

Created attachment 168932 [details]
Patch

[Rik Cabanier]

Please wait for approval from [e-mail], [e-mail], [e-mail], [e-mail] or [e-mail] before submitting, as this patch contains changes to the Chromium public API. See also https://trac.webkit.org/wiki/ChromiumWebKitAPI.

[Yury Semikhatsky]

Comment on attachment 168932 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=168932&action=review

> Source/WebKit/chromium/public/WebDevToolsAgentClient.h:66
> +        virtual size_t visitObject(const void* ptr) = 0;

Please use separate interface for this.

[Ilya Tikhonovsky]

Created attachment 169193 [details]
Patch

[Ilya Tikhonovsky]

Created attachment 169194 [details]
Patch

[Ilya Tikhonovsky]

(In reply to comment #3)
> (From update of attachment 168932 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=168932&action=review
> 
> > Source/WebKit/chromium/public/WebDevToolsAgentClient.h:66
> > +        virtual size_t visitObject(const void* ptr) = 0;
> 
> Please use separate interface for this.

comments addressed

[Yury Semikhatsky]

Comment on attachment 169194 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=169194&action=review

> Source/WebKit/chromium/public/WebDevToolsAgentClient.h:78
> +    virtual void dumpUncountedObjects(const ObjectInfoProvider*) { }

dumpAllocatedObjects?

[Ilya Tikhonovsky]

(In reply to comment #7)
> (From update of attachment 169194 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=169194&action=review
> 
> > Source/WebKit/chromium/public/WebDevToolsAgentClient.h:78
> > +    virtual void dumpUncountedObjects(const ObjectInfoProvider*) { }
> 
> dumpAllocatedObjects?

I am sure that we do not want to print all the allocated objects.
Thus, the current name better suits for this purpose.

[Yury Semikhatsky]

Comment on attachment 169194 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=169194&action=review

> Source/WebKit/chromium/public/WebDevToolsAgentClient.h:72
> +    class ObjectInfoProvider {

Maybe rename ObjectInfoProvider to InstrumentedObjects?

[Vsevolod Vlasov]

Comment on attachment 169194 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=169194&action=review

>>> Source/WebKit/chromium/public/WebDevToolsAgentClient.h:78
>>> +    virtual void dumpUncountedObjects(const ObjectInfoProvider*) { }
>> 
>> dumpAllocatedObjects?
> 
> I am sure that we do not want to print all the allocated objects.
> Thus, the current name better suits for this purpose.

dumpUncountedAllocatedObjects then?

[Vsevolod Vlasov]

Comment on attachment 169194 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=169194&action=review

>> Source/WebKit/chromium/public/WebDevToolsAgentClient.h:72
>> +    class ObjectInfoProvider {
> 
> Maybe rename ObjectInfoProvider to InstrumentedObjects?

AllocatedObjectSizeProvider?

[Ilya Tikhonovsky]

Committed r131869: <http://trac.webkit.org/changeset/131869>