Bug 35570 – WebSocket onmessageerror event handler

• Component: WebCore JavaScript
• Status: RESOLVED FIXED

[Fumitoshi Ukai]

WebSocket onmessageerror event handler is added to report unexpected frames.
http://lists.whatwg.org/pipermail/commit-watchers-whatwg.org/2010/003975.html

[Fumitoshi Ukai]

Created attachment 49877 [details]
Patch

[Alexey Proskuryakov]

Comment on attachment 49877 [details]
Patch

Should we call onmessageerror on frame length overflow for frames 0x80...0xff?

WebSocket::didReceiveMessageError() probably needs a LOG - we have it even in close(), surely an error condition should have one.

Looks good to me otherwise.

[Fumitoshi Ukai]

Created attachment 49973 [details]
Patch

[Fumitoshi Ukai]

(In reply to comment #3)
> Created an attachment (id=49973) [details]
> Patch

Ah, onmessageerror has been changed to onerror
http://lists.whatwg.org/pipermail/commit-watchers-whatwg.org/2010/003979.html

[Alexey Proskuryakov]

Comment on attachment 49973 [details]
Patch

The test depends on WebSocket.onclose patch, correct? I'm still thinking about that idea, not sure if I like it or not.

[Fumitoshi Ukai]

Created attachment 49990 [details]
Patch

[Fumitoshi Ukai]

(In reply to comment #5)
> (From update of attachment 49973 [details])
> The test depends on WebSocket.onclose patch, correct? I'm still thinking about
> that idea, not sure if I like it or not.

no, it doesn't depend on other patch.
Which patch do you mean?

[Alexey Proskuryakov]

OK. I was thinking about bug 35573 (CloseEvent).

[Alexey Proskuryakov]

Comment on attachment 49990 [details]
Patch

> +void WebSocket::didReceiveMessageError()
> +{
> +    LOG(Network, "WebSocket %p didReceiveErrorMessage", this);
> +    if (m_state != OPEN || !scriptExecutionContext())
> +        return;

Can both these conditions really occur? Seems acceptable to be a little defensive, but maybe there should also be ASSERTs if this isn't supposed to happen. Otherwise, these checks can confuse someone about how this code works.

r=me

[Fumitoshi Ukai]

Committed r55573: <http://trac.webkit.org/changeset/55573>

[Fumitoshi Ukai]

(In reply to comment #9)
> (From update of attachment 49990 [details])
> > +void WebSocket::didReceiveMessageError()
> > +{
> > +    LOG(Network, "WebSocket %p didReceiveErrorMessage", this);
> > +    if (m_state != OPEN || !scriptExecutionContext())
> > +        return;
> 
> Can both these conditions really occur? Seems acceptable to be a little
> defensive, but maybe there should also be ASSERTs if this isn't supposed to
> happen. Otherwise, these checks can confuse someone about how this code works.

Ok.

I think scriptExecutionContext() should be non null, but m_state might be CLOSED if script closing the WebSocket and events from network happened same time.

make ASSERT(scriptExecutionContext()) and leave "if (m_state != OPEN)".

> 
> r=me