Evict Service Worker when reading it from disk cache fails.

Evict Service Worker when reading it from disk cache fails.

If reading the SW resources from the disk cache fails, we should evict
the worker or else we will keep getting the same failure.

This patch unsets the worker from the live registration, and then
deletes the registration if the SW was the stored version.

BUG=448003
Committed: https://crrev.com/64e3ffb1c6b85c7d21e63afcbffe85c5a1112eb7
Cr-Commit-Position: refs/heads/master@{#327000}

[falken, 2015-04-27 03:09:41]

falken@chromium.org changed reviewers:
+ kinuko@chromium.org, nhiroki@chromium.org

[falken, 2015-04-27 03:09:41]

This reuses the code to "delete version" when activation fails. Really
activation failing should not trigger the logic (crbug.com/480050) so future
patches will remove/simplify some of that code. I'm not sure we really need to
be concerned about what to do if there's other versions in the pipeline. It
might be better to just force delete the registration and abort any ongoing
registration job. But for now reusing the existing code seems like the simplest
way to see if eviction would drive the start worker failures down.

[nhiroki, 2015-04-27 04:24:25]

lgtm

https://codereview.chromium.org/1098083003/diff/20001/content/browser/service...
File content/browser/service_worker/service_worker_registration.cc (right):

https://codereview.chromium.org/1098083003/diff/20001/content/browser/service...
content/browser/service_worker/service_worker_registration.cc:306: const
scoped_refptr<ServiceWorkerVersion>& version) {
ServiceWorkerRegisterJob::CompleteInternal seems to have a similar logic. Are we
able to replace it with this function and make UnsetVersion() private?

[falken, 2015-04-27 04:58:16]

https://codereview.chromium.org/1098083003/diff/20001/content/browser/service...
File content/browser/service_worker/service_worker_registration.cc (right):

https://codereview.chromium.org/1098083003/diff/20001/content/browser/service...
content/browser/service_worker/service_worker_registration.cc:306: const
scoped_refptr<ServiceWorkerVersion>& version) {
On 2015/04/27 04:24:25, nhiroki wrote:
> ServiceWorkerRegisterJob::CompleteInternal seems to have a similar logic. Are
we
> able to replace it with this function and make UnsetVersion() private?

Yes good point, these are pretty similar. I'll see about refactoring when I
simplify DeleteVersion(), it's following some steps from the old spec that can
be removed.

[falken, 2015-04-27 04:58:37]

The CQ bit was checked by falken@chromium.org

[commit-bot: I haz the power, 2015-04-27 04:58:46]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1098083003/20001

[kinuko, 2015-04-27 05:01:27]

lgtm

[commit-bot: I haz the power, 2015-04-27 05:02:06]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-04-27 05:02:54]

Patchset 2 (id:??) landed as
https://crrev.com/64e3ffb1c6b85c7d21e63afcbffe85c5a1112eb7
Cr-Commit-Position: refs/heads/master@{#327000}

[falken, 2015-04-27 08:01:00]

A revert of this CL (patchset #2 id:20001) has been created in
https://codereview.chromium.org/1056913010/ by falken@chromium.org.

The reason for reverting is: Broke Win 8 Aura content_browsertests

[ RUN      ] ServiceWorkerVersionBrowserTest.ReadResourceFailure
...
[2420:2792:0426/232049:1965296:FATAL:service_worker_registration.cc(44)] Check
failed: ::content::BrowserThread::CurrentlyOn(BrowserThread::IO). Must be called
on Chrome_IOThread; actually called on Unknown Thread.
Backtrace:
	base::debug::StackTrace::StackTrace [0x07FB1081+33]
	logging::LogMessage::~LogMessage [0x07FFD9FF+63]
	content::ServiceWorkerRegistration::~ServiceWorkerRegistration [0x045639F5+293]
	content::ServiceWorkerRegistration::`scalar deleting destructor'
[0x006A087A+26]
	base::RefCounted\u003Ccontent::ServiceWorkerRegistration>::Release
[0x0372C3D2+114]
	scoped_refptr\u003Ccontent::ServiceWorkerRegistration>::Release [0x03D8F94E+14]


http://build.chromium.org/p/chromium.win/builders/Win8%20Aura/builds/29227/st....