Revert "Mixed content iframe checks should use top"

Revert "Mixed content iframe checks should use top"

This reverts commit 147d599fb3afbe23ef39073db6d63abb8d373036.

This patch triggered an assertion in
ChromeRenderViewObserver::allowRunningInsecureContent where the
frame_gurl and origin_host differ.

Also, it breaks almost all ReferrerPolicyTests - not sure that this is a
good thing given that some pages rely on this behavior.

BUG=244210
TBR=abarth@chromium.org

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=153895

[jochen (gone - plz use gerrit), 2013-07-10 14:08:07]

Committed patchset #1 manually as r153895 (presubmit successful).

[abarth-chromium, 2013-07-10 18:02:18]

What do you mean by "it breaks almost all ReferrerPolicyTests" ?

[abarth-chromium, 2013-07-10 18:04:52]

I see.  You mean referrer_policy_browsertest.cc

[jochen (gone - plz use gerrit), 2013-07-10 18:08:24]

On 2013/07/10 18:04:52, abarth wrote:
> I see.  You mean referrer_policy_browsertest.cc

Right, and when I just add the command line switch to allow running insecure
content, I hit the DCHECK in ChromeRenderViewObserver.

Anyway, the tests don't use iframes, and disallowing script execution when
redirecting from https to http and back to https seems to have the potential to
break a lot of sites?

[abarth-chromium, 2013-07-10 18:21:57]

On 2013/07/10 18:08:24, jochen wrote:
> On 2013/07/10 18:04:52, abarth wrote:
> > I see.  You mean referrer_policy_browsertest.cc
> 
> Right, and when I just add the command line switch to allow running insecure
> content, I hit the DCHECK in ChromeRenderViewObserver.
> 
> Anyway, the tests don't use iframes, and disallowing script execution when
> redirecting from https to http and back to https seems to have the potential
to
> break a lot of sites?

Yeah, ok.  I'm just going to go with a straight revert of Tom's CL then instead
of trying to create some new behavior.  We need something we can merge to M29.

Thanks for cleaning up after me.

[abarth-chromium, 2013-07-10 18:32:20]

I've re-opened https://codereview.chromium.org/18104009/ and will continue the
discussion there.