No longer terminating the renderer on bad storage session messages

Changing the behaviour of DOMStorageMessageFilter so that receiving invalid messages does not terminate the renderer.

For background, see remarks in the associated bug.

BUG=69729
TEST=None


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=71990

[Jeff Timanus, 2011-01-20 18:36:41]

Please take a look.

Not much to see.  Jeremy, I followed the suggestions you offered in our off-line
email exhange:

Invalid ID to the following function results in:
* OnStorageAreaId -> just pass back some invalid id (not sure if there
is a constant for this somewhere?)
* Length -> 0
* Key -> null
* GetItem -> null
* SetItem -> <don't set any result>, null
* RemoveItem -> null
* Clear -> false

However, I wasn't certain how to approach dispatch storage event:  I did not see
use of a session id in that routine.

I'm concerned that this change may generate failures when callers receive 'bad'
data.  Are all of the return values suggested above safe?

Jeff

[jam, 2011-01-20 18:43:09]

lgtm

the callers are shutting down in that case, right?  in which case sending them
null rv should be ok, no?

[Jeff Timanus, 2011-01-20 19:11:45]

On 2011/01/20 18:43:09, John Abd-El-Malek wrote:
> lgtm
> 
> the callers are shutting down in that case, right?  in which case sending them
> null rv should be ok, no?

That is my understanding.  The renderer structures should be torn down in
entirety by the time the message is received.

[jam, 2011-01-20 19:22:37]

On Thu, Jan 20, 2011 at 11:11 AM, <twiz@chromium.org> wrote:

> On 2011/01/20 18:43:09, John Abd-El-Malek wrote:
>
>> lgtm
>>
>
>  the callers are shutting down in that case, right?  in which case sending
>> them
>> null rv should be ok, no?
>>
>
> That is my understanding.  The renderer structures should be torn down in
> entirety by the time the message is received.


Yep, so there should be no issues in giving the renderer bad data, since
it's stateless at that point.


>
>
>
>
> http://codereview.chromium.org/6342011/
>

[jam, 2011-01-20 19:29:31]

On Thu, Jan 20, 2011 at 11:22 AM, John Abd-El-Malek <jam@chromium.org>wrote:

>
>
> On Thu, Jan 20, 2011 at 11:11 AM, <twiz@chromium.org> wrote:
>
>> On 2011/01/20 18:43:09, John Abd-El-Malek wrote:
>>
>>> lgtm
>>>
>>
>>  the callers are shutting down in that case, right?  in which case sending
>>> them
>>> null rv should be ok, no?
>>>
>>
>> That is my understanding.  The renderer structures should be torn down in
>> entirety by the time the message is received.
>
>
> Yep, so there should be no issues in giving the renderer bad data, since
> it's stateless at that point.
>

To be clear, I mean it can't persist any state.


>
>
>>
>>
>>
>>
>> http://codereview.chromium.org/6342011/
>>
>
>

[jorlow1, 2011-01-21 09:53:32]

LGTM