Force COM machinery to notice that the path to the DelegateExecute verb handler has changed.

Force COM machinery to notice that the path to the DelegateExecute verb handler has changed.

On update, the path to the local COM server hosting the component changes.  Although the installer updates the path in the registry, it seems that the COM service that runs the handler doesn't notice this.  As a result, Chrome shortcuts stop working after updates.  This change forces COM to notice that the path has changed by attempting to instantiate the class after deleting the registration.  This forces COM to flush its cache, making subsequent uses of the handler by the shell succeed.

BUG=143338
TEST=see bug
TBR=sky@chromium.org

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=154352

[robertshield, 2012-08-30 17:52:43]

looks great, coupla questions

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
File chrome/installer/util/callback_work_item.cc (right):

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
chrome/installer/util/callback_work_item.cc:25: roll_state_ = RS_UNDEFINED;
Should we leave roll_state_ set to RS_FORWARD, which would allow us to alter the
DCHECK in Rollback to ensure that Do() has first been called?

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
File chrome/installer/util/callback_work_item.h (right):

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
chrome/installer/util/callback_work_item.h:12: // A work item that invokes a
callback on Do() and Rollback().
Could you add a comment that gives an example of usage?

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
File chrome/installer/util/work_item_list.h (right):

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
chrome/installer/util/work_item_list.h:14: #include "base/callback.h"
is it possible to use base/callback_forward.h here and in the other #includes
that occur in header files in this CL?

[robertshield, 2012-08-30 17:56:06]

On 2012/08/30 17:52:43, robertshield wrote:
> looks great, 

looks great -> LGTM

coupla questions
> 
>
https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
> File chrome/installer/util/callback_work_item.cc (right):
> 
>
https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
> chrome/installer/util/callback_work_item.cc:25: roll_state_ = RS_UNDEFINED;
> Should we leave roll_state_ set to RS_FORWARD, which would allow us to alter
the
> DCHECK in Rollback to ensure that Do() has first been called?
> 
>
https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
> File chrome/installer/util/callback_work_item.h (right):
> 
>
https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
> chrome/installer/util/callback_work_item.h:12: // A work item that invokes a
> callback on Do() and Rollback().
> Could you add a comment that gives an example of usage?
> 
>
https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
> File chrome/installer/util/work_item_list.h (right):
> 
>
https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
> chrome/installer/util/work_item_list.h:14: #include "base/callback.h"
> is it possible to use base/callback_forward.h here and in the other #includes
> that occur in header files in this CL?

[grt (UTC plus 2), 2012-08-30 18:21:49]

thanks.  TBR'ing @sky for chrome/*.gypi changes.

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
File chrome/installer/util/callback_work_item.cc (right):

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
chrome/installer/util/callback_work_item.cc:25: roll_state_ = RS_UNDEFINED;
On 2012/08/30 17:52:43, robertshield wrote:
> Should we leave roll_state_ set to RS_FORWARD, which would allow us to alter
the
> DCHECK in Rollback to ensure that Do() has first been called?

Discussed in person: doing so would make the DCHECK in IsRollback not work,
which I think is more valuable.  Although I could add new state to detect
Rollback called before Do, it doesn't seem worth it since this would really be
enforcing behavior of WorkItemList, which should be tested independently.

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
File chrome/installer/util/callback_work_item.h (right):

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
chrome/installer/util/callback_work_item.h:12: // A work item that invokes a
callback on Do() and Rollback().
On 2012/08/30 17:52:43, robertshield wrote:
> Could you add a comment that gives an example of usage?

Done.

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
File chrome/installer/util/work_item_list.h (right):

https://chromiumcodereview.appspot.com/10916018/diff/1/chrome/installer/util/...
chrome/installer/util/work_item_list.h:14: #include "base/callback.h"
On 2012/08/30 17:52:43, robertshield wrote:
> is it possible to use base/callback_forward.h here and in the other #includes
> that occur in header files in this CL?

Yes!  All but callback_work_item.h.

[commit-bot: I haz the power, 2012-08-30 21:32:53]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/grt@chromium.org/10916018/7001

[commit-bot: I haz the power, 2012-08-31 02:22:54]

Change committed as 154352