Remove more unnecessary ChromeURLRequestContext members.

Remove more unnecessary ChromeURLRequestContext members.

ProfileIOData is now always deleted on the IO thread. Replaces a use of ChromeURLRequestContextGetter with a ChromeURLDataManagerBackend getter callback. Also gets rid of ExtensionInfoMap from ChromeURLRequestContext by directly passing it into the places that need it. Gets rid of the HTML5 storage objects from ChromeURLRequestContext too.

Adds a workaround for TranslateManager's Profile::GetDefaultRequestContext() use, since after this refactoring (by not requiring ChromeURLRequestContextGetter for ChromeURLDataManagerBackend), it prevents having to initialize the ChromeURLRequestContextGetter in most tests since they don't actually use it. This means |default_request_context_| doesn't always get initialized in BrowserMain(), which causes TranslateManager to crash on startup since it requires |default_request_context_| to be initialized. So we forcibly initialize |default_request_context_| until TranslateManager stops depending on this.

BUG=89396
TEST=none


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=92668

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=92690

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=92804

[willchan no longer on Chromium, 2011-07-02 18:01:52]

+arv,+estade: for webui/
+mkwst: chrome/browser/browsing_data_*

[willchan no longer on Chromium, 2011-07-03 10:50:13]

+mpcomplete for ExtensionInfoMap changes in chrome/browser/extensions/ and
chrome_render_message_filter.*

[Matt Perry, 2011-07-06 18:24:34]

extension bits LGTM

[Evan Stade, 2011-07-06 20:27:54]

I don't really know what's going on here.

http://codereview.chromium.org/7282054/diff/6002/chrome/browser/ui/webui/chro...
File chrome/browser/ui/webui/chrome_url_data_manager.h (right):

http://codereview.chromium.org/7282054/diff/6002/chrome/browser/ui/webui/chro...
chrome/browser/ui/webui/chrome_url_data_manager.h:172: const
base::Callback<ChromeURLDataManagerBackend*(void)> backend_;
comment

[willchan no longer on Chromium, 2011-07-12 13:55:34]

+sky
Sorry Evan, Scott usually reviews those changes and arv rubberstamps them. I've
added Scott now.

Scott: I'm trying to remove the reason for ChromeURLDataManager to go through
the ChromeURLRequestContext to access the ChromeURLDataManagerBackend. Instead,
I'm passing a getter callback directly into ChromeURLDataManager. Note there's
no refcounting here to keep ChromeURLDataManagerBackend alive. I'm counting on
the dependency (which *should* be true, and is definitely a bug if it isn't!)
that ChromeURLDataManager should be killed before the Profile posts the task to
kill the ProfileIOData on the IO thread. Since ProfileIOData owns the
ChromeURLDataManagerBackend, then as long as, on the UI thread, anything
accessing ChromeURLDataManagerBackend (which I believe is only the
ChromeURLDataManager) is killed before the ProfileIOData destruction task gets
posted to the IO thread, we're good. I believe this is already the case.

[sky, 2011-07-12 15:40:51]

LGTM (I reviewed patchset 11, patchset 12 didn't look like it made it up
correctly).

http://codereview.chromium.org/7282054/diff/17001/chrome/browser/ui/webui/chr...
File chrome/browser/ui/webui/chrome_url_data_manager.cc (right):

http://codereview.chromium.org/7282054/diff/17001/chrome/browser/ui/webui/chr...
chrome/browser/ui/webui/chrome_url_data_manager.cc:17: #include
"chrome/browser/profiles/profile.h"
Can we remove this include now?

[Evan Stade, 2011-07-12 16:16:05]

the diffs aren't showing. Just says error: chunk mismatch (?)

[Evan Stade, 2011-07-12 16:17:36]

well, rubber stamp lgtm since it seems Scott's reviewed it

[commit-bot: I haz the power, 2011-07-15 09:52:06]

Change committed as 92668

[commit-bot: I haz the power, 2011-07-15 14:34:52]

Change committed as 92690

[Dmitry Polukhin, 2011-07-18 08:47:23]

http://codereview.chromium.org/7282054/diff/36001/chrome/browser/browser_main.cc
File chrome/browser/browser_main.cc (right):

http://codereview.chromium.org/7282054/diff/36001/chrome/browser/browser_main...
chrome/browser/browser_main.cc:2016: profile->GetRequestContext();
It looks like removing this line broke Chrome on Chrome OS on login screen.
[12795:12795:1542236166900:FATAL:url_fetcher.cc(558)] Check failed:
request_context_getter_. We need an URLRequestContext!
Backtrace:
	base::debug::StackTrace::StackTrace() [0xf3593398]
	logging::LogMessage::~LogMessage() [0xf35b28b1]
	URLFetcher::Core::Start() [0xf412b4d9]
	URLFetcher::Start() [0xf412e3c2]
	TranslateManager::FetchLanguageListFromTranslateServer() [0xf2c1df8f]
	BrowserMain() [0xf2782c0e]
	(anonymous namespace)::RunNamedProcessTypeMain() [0xf26db5f0]
	ChromeMain [0xf26dc012]
	main [0xf26dcbc9]

[Dmitry Polukhin, 2011-07-18 09:06:12]

http://codereview.chromium.org/7282054/diff/36001/chrome/browser/browser_main.cc
File chrome/browser/browser_main.cc (right):

http://codereview.chromium.org/7282054/diff/36001/chrome/browser/browser_main...
chrome/browser/browser_main.cc:2016: profile->GetRequestContext();
On 2011/07/18 08:47:23, Dmitry Polukhin wrote:
> It looks like removing this line broke Chrome on Chrome OS on login screen.
> [12795:12795:1542236166900:FATAL:url_fetcher.cc(558)] Check failed:
> request_context_getter_. We need an URLRequestContext!
> Backtrace:
> 	base::debug::StackTrace::StackTrace() [0xf3593398]
> 	logging::LogMessage::~LogMessage() [0xf35b28b1]
> 	URLFetcher::Core::Start() [0xf412b4d9]
> 	URLFetcher::Start() [0xf412e3c2]
> 	TranslateManager::FetchLanguageListFromTranslateServer() [0xf2c1df8f]
> 	BrowserMain() [0xf2782c0e]
> 	(anonymous namespace)::RunNamedProcessTypeMain() [0xf26db5f0]
> 	ChromeMain [0xf26dc012]
> 	main [0xf26dcbc9]

This line is not guilty. But CL that broke Chrome is right. I reverted it
locally and Chrome is able to show login screen.

[willchan no longer on Chromium, 2011-07-18 10:20:32]

http://codereview.chromium.org/7282054/diff/36001/chrome/browser/browser_main.cc
File chrome/browser/browser_main.cc (right):

http://codereview.chromium.org/7282054/diff/36001/chrome/browser/browser_main...
chrome/browser/browser_main.cc:2016: profile->GetRequestContext();
On 2011/07/18 09:06:12, Dmitry Polukhin wrote:
> On 2011/07/18 08:47:23, Dmitry Polukhin wrote:
> > It looks like removing this line broke Chrome on Chrome OS on login screen.
> > [12795:12795:1542236166900:FATAL:url_fetcher.cc(558)] Check failed:
> > request_context_getter_. We need an URLRequestContext!
> > Backtrace:
> > 	base::debug::StackTrace::StackTrace() [0xf3593398]
> > 	logging::LogMessage::~LogMessage() [0xf35b28b1]
> > 	URLFetcher::Core::Start() [0xf412b4d9]
> > 	URLFetcher::Start() [0xf412e3c2]
> > 	TranslateManager::FetchLanguageListFromTranslateServer() [0xf2c1df8f]
> > 	BrowserMain() [0xf2782c0e]
> > 	(anonymous namespace)::RunNamedProcessTypeMain() [0xf26db5f0]
> > 	ChromeMain [0xf26dc012]
> > 	main [0xf26dcbc9]
> 
> This line is not guilty. But CL that broke Chrome is right. I reverted it
> locally and Chrome is able to show login screen.  

I had encountered that stacktrace earlier in the ChromeOS tests and fixed it by
adding this line. Are you sure you are running a build that includes r92804? I'm
surprised that I could encounter this stacktrace on the bots, fix it, but that
you would still see it. Maybe there's different behavior in the build that's not
covered in tests?

[Dmitry Polukhin, 2011-07-18 11:21:27]

Yes, I'm using TOT Chrome synced on 92834. I have to revert this CL to avoid
crash. You can reproduce it on Linux by building Chrome for Chrome OS and run it
on Linux. 'GYP_DEFINES="chromeos=1 target_arch=ia32" gclient runhooks' and run
'out/Debug/chrome --login-manager'