[Webkit-unassigned] [Bug 50082] [Qt] Use QNetworkAccessManager zerocopy feature from QNetworkReplyHandler

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=50082





--- Comment #11 from Markus Goetz <markus.goetz at nokia.com>  2010-12-11 08:23:51 PST ---
(In reply to comment #10)
> I am sorry that we always nitpick so much in WebKit. I have to admit that both names have issues. ByteBlockQt indicate that it is some generic interface and this is the Qt specific part and QByteBlock has the danger of clashing with future versions of Qt.  :}

Yeah, well, but this one will always have the WebCore namespace :)

> I think for now you can remove the 'capacity' call as no one is calling it?

I might use it later with the image decoders, let's see.

> 
> I also have some questions. So QNetworkAccessManager will internally start changing the QSharedDataPointer<char>? Does this work with realloc? Somehow I would assume to see something like a QList<QByteArray> but I don't. So Qt will still re-alloc internally?

It will not re-allocate.
It goes through this codepath only if you allow it to do so (in the QNetworkRequest) and then only if the server supports it (e.g. content-length header is there). In my measurements 70% of the bytes coming with normal websurfing have that condition.
The QList<QByteArray> thing (actually not that but similar) is used internally in the other 30% of the cases.

Thanks for comments! :)

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.