Bug 133137 – DocumentLoader shouldn't keep ResourceResponses indefinitely unless needed.

• Component: Page Loading
• Status: NEW

[Andreas Kling]

DocumentLoader shouldn't keep ResourceResponses indefinitely unless needed.

[Andreas Kling]

Created attachment 231795 [details]
Patch

[Build Bot]

Comment on attachment 231795 [details]
Patch

Attachment 231795 [details] did not pass mac-wk2-ews (mac-wk2):
Output: http://webkit-queues.appspot.com/results/4714788051484672

New failing tests:
loader/go-back-cached-main-resource.html

[Build Bot]

Created attachment 231799 [details]
Archive of layout-test-results from webkit-ews-16 for mac-mountainlion-wk2

The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews.
Bot: webkit-ews-16  Port: mac-mountainlion-wk2  Platform: Mac OS X 10.8.5

[Geoffrey Garen]

Comment on attachment 231795 [details]
Patch

r=me

[Andreas Kling]

The test is failing because the resource delegate callbacks no longer fire under WK2 as expected. I think we could land this with a platform-specific expectation for WK2.

[Alexey Proskuryakov]

Can you elaborate on why it is ok to not send the delegates in this particular case?

[Andreas Kling]

(In reply to comment #6)
> Can you elaborate on why it is ok to not send the delegates in this particular case?

It's my understanding, from talking with Geoff, that these callbacks exist exclusively for servicing the WebKit1 Mac API, and that they are not needed for WK2.

Is that not the case?

[Alexey Proskuryakov]

It depends on which level the failure occurs at - we could be breaking internal behaviors too. Also, there are a lot of very important WebKit1 clients.

I don't think that "WebKit1 only" is a license to kill. The test seems like a very straightforward one - I expect that the only reason why the patch didn't break dozens of tests is that we don't enable page cache in regression tests by default. But page cache being enabled is the norm outside regression tests.