Refactored code to allow associating workers with multiple renderers.

Refactored code to allow associating workers with multiple renderers.
SharedWorkers now gracefully handle http auth requests after their
initial window has closed.

BUG=27660
TEST=WorkerHttpAuth,SharedWorkerHttpAuth uitests

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=36888

Reverted and reopened due to valgrind failures.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=37365

[jam, 2009-12-22 20:34:10]

http://codereview.chromium.org/509016/diff/2002/3002
File chrome/browser/renderer_host/resource_dispatcher_host.cc (right):

http://codereview.chromium.org/509016/diff/2002/3002#newcode1525
chrome/browser/renderer_host/resource_dispatcher_host.cc:1525: // arbitrary
parent to do so.
Have you run this by anyone on the UI team?  Although a corner case, this seems
confusing to a user (they're using an app, and suddenly an http auth dialog pops
up in another tab).  Will the other tab flash, or will the dialog be modal to
the app?

http://codereview.chromium.org/509016/diff/2002/3006
File chrome/browser/worker_host/worker_document_set.cc (right):

http://codereview.chromium.org/509016/diff/2002/3006#newcode5
chrome/browser/worker_host/worker_document_set.cc:5: #include
"chrome/browser/worker_host/worker_process_host.h"
shouldn't this be worker_document_set.h?  also, set/vector includes should be
first (although I don't think you actually need either of them).

http://codereview.chromium.org/509016/diff/2002/3007
File chrome/browser/worker_host/worker_document_set.h (right):

http://codereview.chromium.org/509016/diff/2002/3007#newcode17
chrome/browser/worker_host/worker_document_set.h:17: class WorkerDocumentSet :
public base::RefCounted<WorkerDocumentSet> {
nit: don't you need to include base/ref_counted so that this header compiles has
all of its includes?

http://codereview.chromium.org/509016/diff/2002/3007#newcode37
chrome/browser/worker_host/worker_document_set.h:37: if (sender() ==
other.sender())
nit: need brace brackets since there's more than 1 line per condition

http://codereview.chromium.org/509016/diff/2002/3007#newcode76
chrome/browser/worker_host/worker_document_set.h:76: // Returns an arbitrary
document
huge nit: period :)

http://codereview.chromium.org/509016/diff/2002/3008
File chrome/browser/worker_host/worker_process_host.cc (right):

http://codereview.chromium.org/509016/diff/2002/3008#newcode533
chrome/browser/worker_host/worker_process_host.cc:533:
parent_iter->render_view_route_id() == render_view_route_id)
nit: need brace brackets here

http://codereview.chromium.org/509016/diff/2002/3011
File chrome/browser/worker_host/worker_service.h (right):

http://codereview.chromium.org/509016/diff/2002/3011#newcode109
chrome/browser/worker_host/worker_service.h:109: // one process per core.
this comment seems off, the function looks like it's checking using a strategy
of one worker per process

http://codereview.chromium.org/509016/diff/2002/3013
File chrome/common/render_messages.h (right):

http://codereview.chromium.org/509016/diff/2002/3013#newcode507
chrome/common/render_messages.h:507: GURL url_;
nit: no trailing underscore for structs (the example above is wrong)

[Andrew T Wilson (Slow), 2010-01-07 20:52:07]

Addressed all your feedback.

http://codereview.chromium.org/509016/diff/2002/3002
File chrome/browser/renderer_host/resource_dispatcher_host.cc (right):

http://codereview.chromium.org/509016/diff/2002/3002#newcode1525
chrome/browser/renderer_host/resource_dispatcher_host.cc:1525: // arbitrary
parent to do so.
On 2009/12/22 20:34:10, John Abd-El-Malek wrote:
> Have you run this by anyone on the UI team?  Although a corner case, this
seems
> confusing to a user (they're using an app, and suddenly an http auth dialog
pops
> up in another tab).  Will the other tab flash, or will the dialog be modal to
> the app?
I have spoken with the UX folks about this in the past, and they are not
particularly concerned about the intersection of HTTP Auth, multiple documents,
and shared workers as it's a fairly obscure case.

In any case, the dialog is not modal and does not cause the tab to flash (at
least not on the mac) - this is a general UX issue, though (there should be some
indication that HTTP Auth dialogs have been displayed on a background tab). I've
filed http://code.google.com/p/chromium/issues/detail?id=31651 for this.

http://codereview.chromium.org/509016/diff/2002/3006
File chrome/browser/worker_host/worker_document_set.cc (right):

http://codereview.chromium.org/509016/diff/2002/3006#newcode5
chrome/browser/worker_host/worker_document_set.cc:5: #include
"chrome/browser/worker_host/worker_process_host.h"
On 2009/12/22 20:34:10, John Abd-El-Malek wrote:
> shouldn't this be worker_document_set.h?  also, set/vector includes should be
> first (although I don't think you actually need either of them).  
Good catch.

http://codereview.chromium.org/509016/diff/2002/3008
File chrome/browser/worker_host/worker_process_host.cc (right):

http://codereview.chromium.org/509016/diff/2002/3008#newcode533
chrome/browser/worker_host/worker_process_host.cc:533:
parent_iter->render_view_route_id() == render_view_route_id)
On 2009/12/22 20:34:10, John Abd-El-Malek wrote:
> nit: need brace brackets here
Added braces, but I'm not sure what the rule is here - I guess it's because the
conditional is multi-line, but the style guide seems vague about this.

[Andrew T Wilson (Slow), 2010-01-20 02:13:30]

Just wanted to make sure this was still on your radar screen (I think you were
on vacation the last couple of weeks).

[jam, 2010-01-21 20:44:10]

lgtm, sorry for the huge delay!  in the future, feel free to send it to someone
else if I'm gone this long so that my vacation doesn't block you :)