extensions: Use base::RunLoop instead of deprecated MessageLoop methods.

extensions: Use base::RunLoop instead of deprecated MessageLoop methods.

Speculative change to fix a crash.

BUG=355083
R=kalman@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=263779

[sadrul, 2014-04-14 04:03:55]

Hi! Is there a notification/observer interface I could use in a test to get
notified of when a call to StartupHelper::InstallFromWebstore() succeeds/fails?

[not at google - send to devlin, 2014-04-14 17:54:48]

lgtm, you'd need to look though the code to see if a relevant notification is
fired, +asargent in case he can think of something.

[sadrul, 2014-04-14 23:22:04]

On 2014/04/14 17:54:48, kalman wrote:
> lgtm, you'd need to look though the code to see if a relevant notification is
> fired,

I didn't find any relevant notifs.

> +asargent in case he can think of something.

[sadrul, 2014-04-15 03:49:39]

On 2014/04/14 23:22:04, sadrul wrote:
> On 2014/04/14 17:54:48, kalman wrote:
> > lgtm, you'd need to look though the code to see if a relevant notification
is
> > fired,
> 
> I didn't find any relevant notifs.
> 
> > +asargent in case he can think of something.

I am going to go ahead with landing this CL so we can merge back the fix for
M35. In the meantime, I will look for a way to add a test for this (if you have
suggestions, please let me know).

[sadrul, 2014-04-15 03:53:13]

Committed patchset #1 manually as r263779.

[sadrul, 2014-04-15 03:54:01]

On 2014/04/15 03:53:13, sadrul wrote:
> Committed patchset #1 manually as r263779.

I had to use --bypass-hooks because it was complaining about the existence of
printf in the file (which is not relevant to this change).

[asargent_no_longer_on_chrome, 2014-04-17 04:27:00]

lgtm

(Sorry for latency - I was out on vacation for several days.)

Feel free to CC me on the bug too (I currently get a 403, presumably it's a
security bug?) if you could use more eyes looking at whatever the problem report
is.

[asargent_no_longer_on_chrome, 2014-04-17 04:46:32]

On 2014/04/17 04:27:00, Antony Sargent wrote:
> lgtm
> 
> (Sorry for latency - I was out on vacation for several days.)
> 
> Feel free to CC me on the bug too (I currently get a 403, presumably it's a
> security bug?) if you could use more eyes looking at whatever the problem
report
> is.

Here's a little more background on what this code is for:

We wanted to have a way to start an install flow of a chrome v2 app from native
code outside the browser. For example, suppose you use another browser as your
primary browser, but want to install some v2 chrome app which feels mostly like
a standalone native app. This code lets callers such as a helper plugin invoke
chrome like this:

chrome --install-from-webstore=<id>

and then instead of running normally, chrome will instead just bring up the
inline install dialog
(https://developer.chrome.com/webstore/inline_installation) from the webstore
and if the user ok's it, proceed to install the app with that id. 

We never really broadcasted the fact that we added this feature or launched a
plugin for other browsers that would use it - it's still in a sort of
experimental state. So unless something funny is going on (eg malicious
developers found this as a workaround for force installing badware extensions or
something), I would expect usage of this flag in the field, and hence hitting
the specific codepath affected by this CL, to be extremely rare.

[sadrul, 2014-04-17 06:20:52]

On 2014/04/17 04:46:32, Antony Sargent wrote:
> On 2014/04/17 04:27:00, Antony Sargent wrote:
> > lgtm
> > 
> > (Sorry for latency - I was out on vacation for several days.)
> > 
> > Feel free to CC me on the bug too (I currently get a 403, presumably it's a
> > security bug?) if you could use more eyes looking at whatever the problem
> report
> > is.
> 
> Here's a little more background on what this code is for:
> 
> We wanted to have a way to start an install flow of a chrome v2 app from
native
> code outside the browser. For example, suppose you use another browser as your
> primary browser, but want to install some v2 chrome app which feels mostly
like
> a standalone native app. This code lets callers such as a helper plugin invoke
> chrome like this:
> 
> chrome --install-from-webstore=<id>
> 
> and then instead of running normally, chrome will instead just bring up the
> inline install dialog
> (https://developer.chrome.com/webstore/inline_installation) from the webstore
> and if the user ok's it, proceed to install the app with that id. 
> 
> We never really broadcasted the fact that we added this feature or launched a
> plugin for other browsers that would use it - it's still in a sort of
> experimental state. So unless something funny is going on (eg malicious
> developers found this as a workaround for force installing badware extensions
or
> something), I would expect usage of this flag in the field, and hence hitting
> the specific codepath affected by this CL, to be extremely rare.

Thanks for the explanation! Indeed, the crashes triggered from this look pretty
rare (I could only find one).