Adjust FrameFetchContext clearing logic when clearing HTMLImportsController

Adjust FrameFetchContext clearing logic when clearing HTMLImportsController

Have the clearing of a Document's HTMLImportsController only clear its
FrameFetchContext if there's no DocumentLoader attached. If a loader is
attached, it is possibly shared amongst multiple Documents and it is
responsible for clearing the fetch context.

This mirrors the condition used during detach().

R=dcheng
BUG=491051
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=195801

[sof, 2015-05-22 16:00:02]

sigbjornf@opera.com changed reviewers:
+ dcheng@chromium.org, japhet@chromium.org, mkwst@chromium.org

[sof, 2015-05-22 16:00:03]

please take a look.

[dcheng, 2015-05-22 16:10:48]

lgtm with a nit

I wish the logic for this was more centralized, but that looks kind of hard to
do easily.

https://codereview.chromium.org/1156663003/diff/1/LayoutTests/fast/loader/jav...
File LayoutTests/fast/loader/javascript-detached-frame-no-crash.html (right):

https://codereview.chromium.org/1156663003/diff/1/LayoutTests/fast/loader/jav...
LayoutTests/fast/loader/javascript-detached-frame-no-crash.html:4:
description("Execute javascript: URL in a replacement document without
crashing.");
Nit: mention something about the child having an import?

[sof, 2015-05-22 16:29:32]

thanks for the quick review.

Loose thinking (on a friday afternoon), but some form of ref counting (deref'ed
on detaches) on the ResourceFetcher might simplify the clearing of the context?

https://codereview.chromium.org/1156663003/diff/1/LayoutTests/fast/loader/jav...
File LayoutTests/fast/loader/javascript-detached-frame-no-crash.html (right):

https://codereview.chromium.org/1156663003/diff/1/LayoutTests/fast/loader/jav...
LayoutTests/fast/loader/javascript-detached-frame-no-crash.html:4:
description("Execute javascript: URL in a replacement document without
crashing.");
On 2015/05/22 16:10:48, dcheng wrote:
> Nit: mention something about the child having an import?

Yes, directly relevant.

[sof, 2015-05-22 16:29:52]

The CQ bit was checked by sigbjornf@opera.com

[sof, 2015-05-22 16:29:53]

The patchset sent to the CQ was uploaded after l-g-t-m from dcheng@chromium.org
Link to the patchset: https://codereview.chromium.org/1156663003/#ps20001
(title: "improve test description")

[commit-bot: I haz the power, 2015-05-22 16:30:08]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1156663003/20001

[Mike West, 2015-05-22 16:53:29]

https://codereview.chromium.org/1156663003/diff/20001/LayoutTests/fast/loader...
File LayoutTests/fast/loader/javascript-detached-frame-no-crash.html (right):

https://codereview.chromium.org/1156663003/diff/20001/LayoutTests/fast/loader...
LayoutTests/fast/loader/javascript-detached-frame-no-crash.html:2: <script
src="../../resources/js-test.js"></script>
Tiny, tiny nit: Use testharness.js next time. Then you can drop the
`-expected.txt` file. Otherwise LGTM 2. :)

[commit-bot: I haz the power, 2015-05-22 18:21:27]

Committed patchset #2 (id:20001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=195801