API::Session should clean up its storage in the network process when destroyed.

• Component: Page Loading
• Status: RESOLVED FIXED

[Andreas Kling]

<rdar://problem/21838764>

[Andreas Kling]

Created attachment 257157 [details]
Proposed patch

[Chris Dumez]

*** Bug 147138 has been marked as a duplicate of this bug. ***

[Alexey Proskuryakov]

Comment on attachment 257157 [details]
Proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=257157&action=review

> Source/WebKit2/ChangeLog:3
> +        API::Session should clean up its storage in the network process when destroyed.

Shouldn't it do the same for WebContent processes?

[Andreas Kling]

Comment on attachment 257157 [details]
Proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=257157&action=review

>> Source/WebKit2/ChangeLog:3
>> +        API::Session should clean up its storage in the network process when destroyed.
> 
> Shouldn't it do the same for WebContent processes?

Hm, for the WebContent processes I only see calls that send:

    Messages::WebProcess::EnsurePrivateBrowsingSession(SessionID::legacyPrivateSessionID())

So they always use the same SessionID, never anything created by WKSessionCreate()

[Alexey Proskuryakov]

Comment on attachment 257157 [details]
Proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=257157&action=review

>>> Source/WebKit2/ChangeLog:3
>>> +        API::Session should clean up its storage in the network process when destroyed.
>> 
>> Shouldn't it do the same for WebContent processes?
> 
> Hm, for the WebContent processes I only see calls that send:
> 
>     Messages::WebProcess::EnsurePrivateBrowsingSession(SessionID::legacyPrivateSessionID())
> 
> So they always use the same SessionID, never anything created by WKSessionCreate()

It looks like WebContent processes get the session ID using WebPage::SetSessionID messages. I don't know why it's done differently for these processes now, would consult with Anders.

[Anders Carlsson]

(In reply to comment #5)
> Comment on attachment 257157 [details]

> It looks like WebContent processes get the session ID using
> WebPage::SetSessionID messages. I don't know why it's done differently for
> these processes now, would consult with Anders.

I think it's less important for the web processes since they have a much shorter lifespan (usually).

Going forward, WKSessionRef should go away in favor of WebsiteDataStore, and it should be impossible to change sessions after the fact (since we haven't supported that feature since Mountain Lion).

[Andreas Kling]

Comment on attachment 257157 [details]
Proposed patch

Thanks for the comment, Anders. cq+'ing.

[WebKit Commit Bot]

Comment on attachment 257157 [details]
Proposed patch

Clearing flags on attachment: 257157

Committed r187115: <http://trac.webkit.org/changeset/187115>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.