Don't attempt to print an error message if there is no console.

Don't attempt to print an error message if there is no console.

frameConsole can return null if the frame associated with DOMWindow
is detached.

BUG=376704

[mkosiba (inactive), 2014-06-02 17:00:00]

Could you PTAL?

The issue is that m_frame->host() is NULL which causes
isCurrentlyDisplayedInFrame to return false, which in turn causes frameConsole()
to be NULL.

If m_frame->host() is NULL that means the frame is detached, so it should be
fine to drop error messages reported for it, right?

[eseidel, 2014-06-02 17:04:12]

We need a test.

[eseidel, 2014-06-02 17:08:07]

Looks like there is a test on the bug, we just need to add it to this patch.

[pfeldman, 2014-06-02 17:25:47]

Can we miss some important messages this way? Actual console is associated with
frame host, so I don't see why it should not be possible to write into detached
frame.

[mkosiba (inactive), 2014-06-03 08:46:28]

On 2014/06/02 17:25:47, pfeldman_ooo wrote:
> Can we miss some important messages this way? Actual console is associated
with
> frame host, so I don't see why it should not be possible to write into
detached
> frame.

This is the bit I was unsure about. The console is there on the frame so we
could equally well fix this by returning the console from
DOMWindow::frameConsole.
It's just that right now the DOMWindow::frameConsole() method is returning NULL
because of an early return.
The early return was added together with the frameConsole method when
PageConsole was
split out from Console. Pavel, looks like the original patchset was witten by
you
(if I'm reading https://bugs.webkit.org/show_bug.cgi?id=111578 correctly). Do
you
remember by any chance why did you add the early return? This is what the code
looked like when the method was added:

FrameConsole* DOMWindow::frameConsole() const
{
    if (!isCurrentlyDisplayedInFrame())
        return 0;
    return m_frame->page() ? m_frame->page()->console() : 0;
}

[mkosiba (inactive), 2014-06-03 09:55:34]

On 2014/06/02 17:08:07, eseidel wrote:
> Looks like there is a test on the bug, we just need to add it to this patch.

Done.

[eseidel, 2014-06-07 13:55:53]

i'm confused as to what error message this test even makes to trigger the crash?

[mkosiba (inactive), 2014-06-09 16:50:34]

On 2014/06/07 13:55:53, eseidel (OOO until June 16) wrote:
> i'm confused as to what error message this test even makes to trigger the
crash?

It gets triggered from the second printNavigationErrorMessage call from
Document::canNavigate

[mkosiba (inactive), 2014-06-09 16:50:51]

On 2014/06/09 16:50:34, mkosiba wrote:
> On 2014/06/07 13:55:53, eseidel (OOO until June 16) wrote:
> > i'm confused as to what error message this test even makes to trigger the
> crash?
> 
> It gets triggered from the second printNavigationErrorMessage call from
> Document::canNavigate

the line is:
    printNavigationErrorMessage(toLocalFrameTemporary(targetFrame), url(), "The
frame attempting navigation is neither same-origin with the target, nor is it
the target's parent or opener.");

[pfeldman, 2014-06-16 17:25:48]

> FrameConsole* DOMWindow::frameConsole() const
> {
>     if (!isCurrentlyDisplayedInFrame())
>         return 0;
>     return m_frame->page() ? m_frame->page()->console() : 0;
> }

I think I was reviewing it, not authoring. In either case, I don't see why this
check is necessary.

[mkosiba (inactive), 2014-06-26 17:03:04]

On 2014/06/16 17:25:48, pfeldman wrote:
> > FrameConsole* DOMWindow::frameConsole() const
> > {
> >     if (!isCurrentlyDisplayedInFrame())
> >         return 0;
> >     return m_frame->page() ? m_frame->page()->console() : 0;
> > }
> 
> I think I was reviewing it, not authoring. In either case, I don't see why
this
> check is necessary.

Done, PTAL