I actually don't deal much with the debugger protocol, as in Chrome
Developer Tools we use our own.

On Wed, Dec 22, 2010 at 11:50, Søren Gjesse <[email protected]> wrote:
> Ryan,
> This sounds like a reasonable request. The new heap profiler can generate a
> JSON representation of a heap snapshot, so that would fit straight into the
> debugger protocol. However this JSON string have a size which could easily
> be the same size as the heap, so using it directly is probably not the
> right approach.

Yes, I plan to address that. Generating this string doesn't cost
anything on the sender side. As for the receiver -- the snapshot
consists of two major parts: the heap graph, and strings, the former
surely needs to be deserialized, as it needs to be processed as a
whole, but strings deserialization can be deferred, and they can be
injected into receiver's JS heap as external strings.

> To get this working a number of protocol requests should be designed, some
> to make snapshots and sone to retrieve this information incrementally. We
> should also consider if the objects in the snapshots can in some way be
> related to the mirror objects used by the debugger.

Another approach may be not to send the whole snapshot, but instead
retrieve it remotely via the v8-profiler.h API.

> Regards,
> Søren
>
> On Wed, Dec 22, 2010 at 09:11, Ryan Dahl <[email protected]> wrote:
>>
>> Is there a desire to have this or a reason not to have it?
>>
>> --
>> v8-users mailing list
>> [email protected]
>> http://groups.google.com/group/v8-users
>
> --
> v8-users mailing list
> [email protected]
> http://groups.google.com/group/v8-users

-- 
v8-users mailing list
[email protected]
http://groups.google.com/group/v8-users

Reply via email to