On 2009/08/04 21:51:20, Peter Rybin wrote:

Adding a break command the the debugger agent protocol is definitely
something which is needed. However I think this should be handled as
part of the remote debugger protocol and not through a special command.
The v8:Debug::SendCommand() should stay as a JSON interface. With your
suggestion calling v8:Debug::SendCommand() with the command "break" will
be equivalent to calling v8::Debug::DebugBreak().

I suggest adding a header to the protocol for signalling "out-of-band"
stuff like break, e.g. this message would cause a break.

Debug-Break: true
Content-Length: 0

This would require adding handling of other headers than Content-Length
in the protocol (currently they are just printed).

We also need to figure out whether there should be a "forced" break,
where the embedding application is requested to execute some JavaScript
(e.g. 'void 0') to ensure that the break is "immediate" as opposed to
"when JavaScript runs". This will required some extra facilities for
calling back to the embedder as the debugger agent cannot safely
initiate execution of JavaScript due to threading issues.

http://codereview.chromium.org/160605

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

Reply via email to