Introduce a new 'all_frames' property to content scripts and

Introduce a new 'all_frames' property to content scripts and
default it to false. This should improve performance in
sites that use frames and elimiate confusion, since in most
cases developers *don't* expect content scripts to match
frames.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=33035

[Erik does not do reviews, 2009-11-20 23:14:45]

http://codereview.chromium.org/412008/diff/2004/3019
File chrome/common/extensions/docs/static/content_scripts.html (right):

http://codereview.chromium.org/412008/diff/2004/3019#newcode133
Line 133: <td><em>Optional.</em></td>
move the </td> to the end, it looks like the compiler freaked out when it saw
this (verify on the compiled page)

http://codereview.chromium.org/412008/diff/2004/3028
File chrome/test/data/extensions/api_test/content_script_all_frames/test.html
(right):

http://codereview.chromium.org/412008/diff/2004/3028#newcode22
Line 22: if (num_all_frames_messages == 2 && num_top_frame_only_messages == 1) {
This test doesn't feel that reliable to me.  If more messages are still coming
at this point, you wouldn't know it because you exit immediately.  Maybe you
need to have a second message in order to guarantee that the first message was
completely processed?

http://codereview.chromium.org/412008/diff/2004/3031
File chrome/test/data/extensions/test_file_with_iframe.html (right):

http://codereview.chromium.org/412008/diff/2004/3031#newcode3
Line 3: <iframe src="test_file.html"></iframe>
it looks like this file isn't in the CL

[Aaron Boodman, 2009-11-21 07:49:59]

http://codereview.chromium.org/412008/diff/2004/3028
File chrome/test/data/extensions/api_test/content_script_all_frames/test.html
(right):

http://codereview.chromium.org/412008/diff/2004/3028#newcode22
Line 22: if (num_all_frames_messages == 2 && num_top_frame_only_messages == 1) {
On 2009/11/20 23:14:45, Erik Kay wrote:
> This test doesn't feel that reliable to me.  If more messages are still coming
> at this point, you wouldn't know it because you exit immediately.  Maybe you
> need to have a second message in order to guarantee that the first message was
> completely processed?

I understand what you mean, but I don't see a general way to solve this problem.

We expect three messages and wait for three. Even if we had a fourth "i'm all
done now" message, we could never know that something wouldn't have gone haywire
after that causing more messages to be sent.

I think those are bugs that we will have to discover and write specific tests
for.

http://codereview.chromium.org/412008/diff/2004/3031
File chrome/test/data/extensions/test_file_with_iframe.html (right):

http://codereview.chromium.org/412008/diff/2004/3031#newcode3
Line 3: <iframe src="test_file.html"></iframe>
On 2009/11/20 23:14:45, Erik Kay wrote:
> it looks like this file isn't in the CL

The file already exists in the repro
(chrome/test/data/extensions/test_file.html).

[Erik does not do reviews, 2009-11-23 15:47:12]

http://codereview.chromium.org/412008/diff/2004/3028
File chrome/test/data/extensions/api_test/content_script_all_frames/test.html
(right):

http://codereview.chromium.org/412008/diff/2004/3028#newcode22
chrome/test/data/extensions/api_test/content_script_all_frames/test.html:22: if
(num_all_frames_messages == 2 && num_top_frame_only_messages == 1) {
On 2009/11/21 07:49:59, Aaron Boodman wrote:
> On 2009/11/20 23:14:45, Erik Kay wrote:
> > This test doesn't feel that reliable to me.  If more messages are still
coming
> > at this point, you wouldn't know it because you exit immediately.  Maybe you
> > need to have a second message in order to guarantee that the first message
was
> > completely processed?
> 
> I understand what you mean, but I don't see a general way to solve this
problem.
> 
> We expect three messages and wait for three. Even if we had a fourth "i'm all
> done now" message, we could never know that something wouldn't have gone
haywire
> after that causing more messages to be sent.
> 
> I think those are bugs that we will have to discover and write specific tests
> for.
What I meant was that if you sent a second separate message from the background
page, it would always be delivered after the first message, right?  My hope was
that should mean that the reply would get delivered after all of the other
replies as well.

I'm fine with punting on this for now, just add a comment warning about the
issue.