Quit BlinkTestRunner::TestFinished early if main frame is not local.

Quit BlinkTestRunner::TestFinished early if main frame is not local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout from BlinkTestRunner::CaptureDump (which targets
the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

Note - this CL prevents a crash in a future CL
(https://crrev.com/1750063002), by enforcing the assumption described
above (and exiting early if the assumption doesn't hold).  OTOH, this CL
does NOT actually cause ShellViewHostMsg_TestFinishedInSecondaryRenderer
to be delivered to the browser in all the cases - there are extra issues
that prevent that from happening (these issues are addressed in a later
CL proposed at https://crrev.com/1765623003/).  Not delivering the
message is harmless and actually prevents test failures at this
point (when runtime test flags are not yet replicated - until we land
https://crrev.com/1715573002/).

BUG=587175
Committed: https://crrev.com/f6b148302e76a6450d87ea49ef9e6f9836ca2639
Cr-Commit-Position: refs/heads/master@{#379870}

[Łukasz Anforowicz, 2016-03-01 20:07:27]

Description was changed from

==========
Ensure that BlinkTestRunner::TestFinished is only when main frame is local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.  This
assumption
is needed for test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout.  This assumption accidentally holds true today, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message which
means
that BlinkTestRunner::is_main_window_ is set to false in OOPIFs (even if they
are in the main window).

This CL makes sure that the assumption above will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to OOPIFs in
the future.  This is done, by having OOPIFs ask the browser to forward
TestFinished to the BlinkTestRunner where main frame is local (similarily to
how this was already asked from renderers associated with secondary test
windows).  Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

This CL means that presence of OOPIFs can cause different browser/renderer
messages
to be exchanged between BlinkTestController and BlinkTestRunner.  This is
unfortunate,
because it means that synchronization done by the test in default running mode
might be insufficient with --site-per-process mode.  Ideally, we would
consistently
forward TestFinished processing to the main frame's renderer, regardless of
whether
OOPIFs are active or not (i.e. basing the forwarding decision on whether the
end of test was trigerred by the main frame vs by a child frame). 
Unfortunately,
BlinkTestRunner::TestFinished is not capable of detecting which frame triggered
end of test (sometimes the test is triggered by a specific frame calling
testRunner.notifyDone, but sometimes the test is triggered in a frame-agnostic
way when the queue in test_runner::TestRunner::WorkQueue gets empty).

BUG=587175
==========

to

==========
Ensure that BlinkTestRunner::TestFinished is only called when main frame is
local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.  This
assumption
is needed for test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout.  This assumption accidentally holds true today, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message which
means
that BlinkTestRunner::is_main_window_ is set to false in OOPIFs (even if they
are in the main window).

This CL makes sure that the assumption above will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to OOPIFs in
the future.  This is done, by having OOPIFs ask the browser to forward
TestFinished to the BlinkTestRunner where main frame is local (similarily to
how this was already asked from renderers associated with secondary test
windows).  Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

This CL means that presence of OOPIFs can cause different browser/renderer
messages
to be exchanged between BlinkTestController and BlinkTestRunner.  This is
unfortunate,
because it means that synchronization done by the test in default running mode
might be insufficient with --site-per-process mode.  Ideally, we would
consistently
forward TestFinished processing to the main frame's renderer, regardless of
whether
OOPIFs are active or not (i.e. basing the forwarding decision on whether the
end of test was trigerred by the main frame vs by a child frame). 
Unfortunately,
BlinkTestRunner::TestFinished is not capable of detecting which frame triggered
end of test (sometimes the test is triggered by a specific frame calling
testRunner.notifyDone, but sometimes the test is triggered in a frame-agnostic
way when the queue in test_runner::TestRunner::WorkQueue gets empty).

BUG=587175
==========

[Łukasz Anforowicz, 2016-03-02 23:44:48]

Description was changed from

==========
Ensure that BlinkTestRunner::TestFinished is only called when main frame is
local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.  This
assumption
is needed for test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout.  This assumption accidentally holds true today, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message which
means
that BlinkTestRunner::is_main_window_ is set to false in OOPIFs (even if they
are in the main window).

This CL makes sure that the assumption above will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to OOPIFs in
the future.  This is done, by having OOPIFs ask the browser to forward
TestFinished to the BlinkTestRunner where main frame is local (similarily to
how this was already asked from renderers associated with secondary test
windows).  Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

This CL means that presence of OOPIFs can cause different browser/renderer
messages
to be exchanged between BlinkTestController and BlinkTestRunner.  This is
unfortunate,
because it means that synchronization done by the test in default running mode
might be insufficient with --site-per-process mode.  Ideally, we would
consistently
forward TestFinished processing to the main frame's renderer, regardless of
whether
OOPIFs are active or not (i.e. basing the forwarding decision on whether the
end of test was trigerred by the main frame vs by a child frame). 
Unfortunately,
BlinkTestRunner::TestFinished is not capable of detecting which frame triggered
end of test (sometimes the test is triggered by a specific frame calling
testRunner.notifyDone, but sometimes the test is triggered in a frame-agnostic
way when the queue in test_runner::TestRunner::WorkQueue gets empty).

BUG=587175
==========

to

==========
Ensure that BlinkTestRunner::TestFinished is only called when main frame is
local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout.  This assumption accidentally holds true today,
because today OOPIFs do not receive ShellViewMsg_SetTestConfiguration
message which means that BlinkTestRunner::is_main_window_ is set to
false in OOPIFs (even if they actually *are* in the main window).

This CL makes sure that the assumption above will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

This CL means that presence of OOPIFs can cause different
browser/renderer messages to be exchanged between BlinkTestController
and BlinkTestRunner.  This is unfortunate, because it means that
synchronization done by the test in default running mode might be
insufficient with --site-per-process mode.  Ideally, we would
consistently forward TestFinished processing to the main frame's
renderer, regardless of whether OOPIFs are active or not (i.e. basing
the forwarding decision on whether the end of test was trigerred by the
main frame vs by a child frame).  Unfortunately,
BlinkTestRunner::TestFinished is not capable of detecting which frame
triggered end of test (sometimes the test is triggered by a specific
frame calling testRunner.notifyDone, but sometimes the test is triggered
in a frame-agnostic way when the queue in
test_runner::TestRunner::WorkQueue gets empty).

BUG=587175
==========

[Łukasz Anforowicz, 2016-03-03 19:04:35]

Description was changed from

==========
Ensure that BlinkTestRunner::TestFinished is only called when main frame is
local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout.  This assumption accidentally holds true today,
because today OOPIFs do not receive ShellViewMsg_SetTestConfiguration
message which means that BlinkTestRunner::is_main_window_ is set to
false in OOPIFs (even if they actually *are* in the main window).

This CL makes sure that the assumption above will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

This CL means that presence of OOPIFs can cause different
browser/renderer messages to be exchanged between BlinkTestController
and BlinkTestRunner.  This is unfortunate, because it means that
synchronization done by the test in default running mode might be
insufficient with --site-per-process mode.  Ideally, we would
consistently forward TestFinished processing to the main frame's
renderer, regardless of whether OOPIFs are active or not (i.e. basing
the forwarding decision on whether the end of test was trigerred by the
main frame vs by a child frame).  Unfortunately,
BlinkTestRunner::TestFinished is not capable of detecting which frame
triggered end of test (sometimes the test is triggered by a specific
frame calling testRunner.notifyDone, but sometimes the test is triggered
in a frame-agnostic way when the queue in
test_runner::TestRunner::WorkQueue gets empty).

BUG=587175
==========

to

==========
Ensure that BlinkTestRunner::TestFinished is only called when main frame is
local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for test_runner::TestRunner::CheckResponseMimeType
and for the call to test_runner::DumpLayout from BlinkTestRunner::CaptureDump
(which targets the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

BUG=587175
==========

[Łukasz Anforowicz, 2016-03-03 19:05:02]

lukasza@chromium.org changed reviewers:
+ jochen@chromium.org

[Łukasz Anforowicz, 2016-03-03 19:05:03]

jochen@, can you please take a look?

[jochen (gone - plz use gerrit), 2016-03-04 12:38:52]

https://codereview.chromium.org/1746393002/diff/20001/content/shell/renderer/...
File content/shell/renderer/layout_test/blink_test_runner.cc (right):

https://codereview.chromium.org/1746393002/diff/20001/content/shell/renderer/...
content/shell/renderer/layout_test/blink_test_runner.cc:614: if
(!is_main_window_ || !render_view()->GetMainRenderFrame()) {
ah, so you consider all OOPIF of the main window to be part of the main window?

what if you have a test which opens a window in a different domain, and has an
iframe from that different domain in the main window? won't the other window and
the oopif end up in the same process then?

[Łukasz Anforowicz, 2016-03-04 20:41:41]

https://codereview.chromium.org/1746393002/diff/20001/content/shell/renderer/...
File content/shell/renderer/layout_test/blink_test_runner.cc (right):

https://codereview.chromium.org/1746393002/diff/20001/content/shell/renderer/...
content/shell/renderer/layout_test/blink_test_runner.cc:614: if
(!is_main_window_ || !render_view()->GetMainRenderFrame()) {
On 2016/03/04 12:38:52, jochen wrote:
> ah, so you consider all OOPIF of the main window to be part of the main
window?

Correct, except to be precise s/OOPIF/RenderViews/ (i.e. each rendererer
containing a frame from the main window will also have a RenderView +
BlinkTestRunner associated with the window/page-level stuff).

> what if you have a test which opens a window in a different domain, and has an
> iframe from that different domain in the main window? won't the other window
and
> the oopif end up in the same process then?

They will be in the same process, but they will have 2 separate RenderViews. 
Creation of both RenderViews will go through
LayoutTestContentRendererClient::WebTestProxyCreated, so they both will have an
associated BlinkTestRunner.  OTOH, only one of these BlinkTestRunners will
receive a SetTestConfiguration (or ReplicateTestConfiguration) message (only the
one associated with the RenderView for the test's main window).  So I think this
is ok?

[Łukasz Anforowicz, 2016-03-04 20:51:19]

https://codereview.chromium.org/1746393002/diff/20001/content/shell/renderer/...
File content/shell/renderer/layout_test/blink_test_runner.cc (right):

https://codereview.chromium.org/1746393002/diff/20001/content/shell/renderer/...
content/shell/renderer/layout_test/blink_test_runner.cc:614: if
(!is_main_window_ || !render_view()->GetMainRenderFrame()) {
On 2016/03/04 20:41:41, Łukasz Anforowicz wrote:
> On 2016/03/04 12:38:52, jochen wrote:
> > ah, so you consider all OOPIF of the main window to be part of the main
> window?
> 
> Correct, except to be precise s/OOPIF/RenderViews/ (i.e. each rendererer
> containing a frame from the main window will also have a RenderView +
> BlinkTestRunner associated with the window/page-level stuff).
> 
> > what if you have a test which opens a window in a different domain, and has
an
> > iframe from that different domain in the main window? won't the other window
> and
> > the oopif end up in the same process then?
> 
> They will be in the same process, but they will have 2 separate RenderViews. 
> Creation of both RenderViews will go through
> LayoutTestContentRendererClient::WebTestProxyCreated, so they both will have
an
> associated BlinkTestRunner.  OTOH, only one of these BlinkTestRunners will
> receive a SetTestConfiguration (or ReplicateTestConfiguration) message (only
the
> one associated with the RenderView for the test's main window).  So I think
this
> is ok?

i.e. only one of the BlinkTestRunner's will have |is_main_window_| set to true.

[Łukasz Anforowicz, 2016-03-07 19:22:36]

Description was changed from

==========
Ensure that BlinkTestRunner::TestFinished is only called when main frame is
local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for test_runner::TestRunner::CheckResponseMimeType
and for the call to test_runner::DumpLayout from BlinkTestRunner::CaptureDump
(which targets the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

BUG=587175
==========

to

==========
Quit BlinkTestRunner::TestFinished early if main frame is not local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout from BlinkTestRunner::CaptureDump (which targets
the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

Note - this CL prevents a crash in a future CL
(https://crrev.com/1750063002), by enforcing the assumption described
above (and exiting early if the assumption doesn't hold).  OTOH, this CL
does NOT actually cause ShellViewHostMsg_TestFinishedInSecondaryRenderer
to be delivered to the browser in all the cases - there are extra issues
that prevent that from happening (these issues are addressed in a later
CL proposed at https://crrev.com/1765623003/).  Not delivering the
message is not harmless and actually prevents test failures at this
point (when runtime test flags are not yet delivered - until we land
https://crrev.com/1715573002/).

BUG=587175
==========

[Łukasz Anforowicz, 2016-03-07 19:22:47]

Description was changed from

==========
Quit BlinkTestRunner::TestFinished early if main frame is not local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout from BlinkTestRunner::CaptureDump (which targets
the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

Note - this CL prevents a crash in a future CL
(https://crrev.com/1750063002), by enforcing the assumption described
above (and exiting early if the assumption doesn't hold).  OTOH, this CL
does NOT actually cause ShellViewHostMsg_TestFinishedInSecondaryRenderer
to be delivered to the browser in all the cases - there are extra issues
that prevent that from happening (these issues are addressed in a later
CL proposed at https://crrev.com/1765623003/).  Not delivering the
message is not harmless and actually prevents test failures at this
point (when runtime test flags are not yet delivered - until we land
https://crrev.com/1715573002/).

BUG=587175
==========

to

==========
Quit BlinkTestRunner::TestFinished early if main frame is not local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout from BlinkTestRunner::CaptureDump (which targets
the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

Note - this CL prevents a crash in a future CL
(https://crrev.com/1750063002), by enforcing the assumption described
above (and exiting early if the assumption doesn't hold).  OTOH, this CL
does NOT actually cause ShellViewHostMsg_TestFinishedInSecondaryRenderer
to be delivered to the browser in all the cases - there are extra issues
that prevent that from happening (these issues are addressed in a later
CL proposed at https://crrev.com/1765623003/).  Not delivering the
message is not harmless and actually prevents test failures at this
point (when runtime test flags are not yet delivered - until we land
https://crrev.com/1715573002/).

BUG=587175
==========

[Łukasz Anforowicz, 2016-03-07 22:21:59]

Description was changed from

==========
Quit BlinkTestRunner::TestFinished early if main frame is not local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout from BlinkTestRunner::CaptureDump (which targets
the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

Note - this CL prevents a crash in a future CL
(https://crrev.com/1750063002), by enforcing the assumption described
above (and exiting early if the assumption doesn't hold).  OTOH, this CL
does NOT actually cause ShellViewHostMsg_TestFinishedInSecondaryRenderer
to be delivered to the browser in all the cases - there are extra issues
that prevent that from happening (these issues are addressed in a later
CL proposed at https://crrev.com/1765623003/).  Not delivering the
message is not harmless and actually prevents test failures at this
point (when runtime test flags are not yet delivered - until we land
https://crrev.com/1715573002/).

BUG=587175
==========

to

==========
Quit BlinkTestRunner::TestFinished early if main frame is not local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout from BlinkTestRunner::CaptureDump (which targets
the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

Note - this CL prevents a crash in a future CL
(https://crrev.com/1750063002), by enforcing the assumption described
above (and exiting early if the assumption doesn't hold).  OTOH, this CL
does NOT actually cause ShellViewHostMsg_TestFinishedInSecondaryRenderer
to be delivered to the browser in all the cases - there are extra issues
that prevent that from happening (these issues are addressed in a later
CL proposed at https://crrev.com/1765623003/).  Not delivering the
message is harmless and actually prevents test failures at this
point (when runtime test flags are not yet replicated - until we land
https://crrev.com/1715573002/).

BUG=587175
==========

[jochen (gone - plz use gerrit), 2016-03-08 16:23:49]

lgtm

[Łukasz Anforowicz, 2016-03-08 17:33:23]

The CQ bit was checked by lukasza@chromium.org

[commit-bot: I haz the power, 2016-03-08 17:33:42]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1746393002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1746393002/40001

[commit-bot: I haz the power, 2016-03-08 18:31:07]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-03-08 18:32:43]

Description was changed from

==========
Quit BlinkTestRunner::TestFinished early if main frame is not local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout from BlinkTestRunner::CaptureDump (which targets
the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

Note - this CL prevents a crash in a future CL
(https://crrev.com/1750063002), by enforcing the assumption described
above (and exiting early if the assumption doesn't hold).  OTOH, this CL
does NOT actually cause ShellViewHostMsg_TestFinishedInSecondaryRenderer
to be delivered to the browser in all the cases - there are extra issues
that prevent that from happening (these issues are addressed in a later
CL proposed at https://crrev.com/1765623003/).  Not delivering the
message is harmless and actually prevents test failures at this
point (when runtime test flags are not yet replicated - until we land
https://crrev.com/1715573002/).

BUG=587175
==========

to

==========
Quit BlinkTestRunner::TestFinished early if main frame is not local.

content::BlinkTestRunner::TestFinished assumes that main frame is local.
This assumption only accidentally holds true today when running layout
tests with --site-per-process flag - it holds true with OOPIFs, because
today OOPIFs do not receive ShellViewMsg_SetTestConfiguration message
which means that BlinkTestRunner::is_main_window_ is set to false in
OOPIFs (even if they actually *are* in the main window).

This assumption is needed for
test_runner::TestRunner::CheckResponseMimeType and for the call to
test_runner::DumpLayout from BlinkTestRunner::CaptureDump (which targets
the main frame only).

This CL makes sure that the assumption will hold true even if an
equivalent of ShellViewMsg_SetTestConfiguration message is sent to
OOPIFs in the future (i.e. after crrev.com/1750063002).  This is done,
by having OOPIFs ask the browser to forward TestFinished to the
BlinkTestRunner where main frame is local (similarily to how this was
already asked from renderers associated with secondary test windows).
Because of this change, ...SecondaryWindow... is replaced with
...SecondaryRenderer... where applicable.

Note - this CL prevents a crash in a future CL
(https://crrev.com/1750063002), by enforcing the assumption described
above (and exiting early if the assumption doesn't hold).  OTOH, this CL
does NOT actually cause ShellViewHostMsg_TestFinishedInSecondaryRenderer
to be delivered to the browser in all the cases - there are extra issues
that prevent that from happening (these issues are addressed in a later
CL proposed at https://crrev.com/1765623003/).  Not delivering the
message is harmless and actually prevents test failures at this
point (when runtime test flags are not yet replicated - until we land
https://crrev.com/1715573002/).

BUG=587175

Committed: https://crrev.com/f6b148302e76a6450d87ea49ef9e6f9836ca2639
Cr-Commit-Position: refs/heads/master@{#379870}
==========

[commit-bot: I haz the power, 2016-03-08 18:32:44]

Patchset 3 (id:??) landed as
https://crrev.com/f6b148302e76a6450d87ea49ef9e6f9836ca2639
Cr-Commit-Position: refs/heads/master@{#379870}