Add ResourceRequest target types for worker and shared workers....

Ground work for making appcaches work in workers.

Add ResourceRequest target types for worker and shared worker resources. This is the first patch of a two-sided change that involves WebKit API changes which haven't been made yet.

BUG=39368
TEST=none



Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=45643

[michaeln, 2010-04-24 02:00:56]

Please take a look

[Andrew T Wilson (Slow), 2010-04-25 17:13:15]

LGTM

http://codereview.chromium.org/1666002/diff/5001/6002
File webkit/glue/resource_type.h (right):

http://codereview.chromium.org/1666002/diff/5001/6002#newcode23
webkit/glue/resource_type.h:23: SHARED_WORKER,   // the main resource of a
shared worker.
When workers perform XHR or load scripts like importScripts() I'm guessing these
use SUB_RESOURCE type, then? So the only time WORKER/SHARED_WORKER is used is
when loading the initial script resource via the Worker()/SharedWorker()
constructors? These would go away if/when we ever move script loading to the
worker, right?

(I think that's the correct implementation - just making sure I understand it).

[michaeln, 2010-04-26 21:51:53]

http://codereview.chromium.org/1666002/diff/5001/6002
File webkit/glue/resource_type.h (right):

http://codereview.chromium.org/1666002/diff/5001/6002#newcode23
webkit/glue/resource_type.h:23: SHARED_WORKER,   // the main resource of a
shared worker.
On 2010/04/25 17:13:15, Andrew T Wilson wrote:
> When workers perform XHR or load scripts like importScripts() I'm guessing
these
> use SUB_RESOURCE type, then?

I think right now all resources loaded by workers get tagged as SUB_RESOURCEs
becauses there is no thread marshaling for the 'type' data member. When I do the
webkit/webcore patch we'll have a choice about how to tag those reource
requests.

The most accurate 'type' for importScripts() seems like the SCRIPT_RESOURCE
type. Not posted for review yet but here's a preview of the webcore changes.
http://codereview.chromium.org/1521038/show


> So the only time WORKER/SHARED_WORKER is used is
> when loading the initial script resource via the Worker()/SharedWorker()
> constructors? These would go away if/when we ever move script loading to the
> worker, right?

Probably makes sense to have those 'types' regardless of how/where the request
for those resources are made. If the callsite to load the main resource for a
worker moved to the worker process itself (not gonna happen any time soon i
think), we'd use the appropriate type at the new callsite.