PasswordController initialization: WebStateObserverBridge to the end

PasswordController initialization: WebStateObserverBridge to the end

The associated bug has a crash where a WebStateObserverBridge method of
PasswordController dereferences a null pointer. The block where the dereference
happens is:

    if (weakSelf) {
      weakSelf.get()->passwordManager_->OnPasswordFormSubmitted(
          weakSelf.get()->passwordManagerDriver_.get(), form);
    }

While the most likely cause of that bug is being addressed in
https://codereview.chromium.org/1941363002/, the initialization order in
PasswordController's -initWithWebState is also suspicious: PasswordController
starts observing the WebState before it creates the supporting objects like
passwordManager_.

This CL changes the initialisation order to only create the bridge once the two
pointers are non-null.

BUG=608090
Committed: https://crrev.com/bcd8636b075e7bcf9f5861f981695236dd533baf
Cr-Commit-Position: refs/heads/master@{#392588}

[vabr (Chromium), 2016-05-04 14:53:55]

vabr@chromium.org changed reviewers:
+ vasilii@chromium.org

[vabr (Chromium), 2016-05-04 14:53:55]

Hi Vasilii,

Does this make sense to you?

Cheers,
Vaclav

[vasilii, 2016-05-10 12:38:47]

https://codereview.chromium.org/1950983002/diff/20001/ios/chrome/browser/pass...
File ios/chrome/browser/passwords/password_controller.mm (right):

https://codereview.chromium.org/1950983002/diff/20001/ios/chrome/browser/pass...
ios/chrome/browser/passwords/password_controller.mm:244: web::WebState*
webState_;               // weak
Why is the new member required?

[vabr (Chromium), 2016-05-10 12:48:47]

https://codereview.chromium.org/1950983002/diff/20001/ios/chrome/browser/pass...
File ios/chrome/browser/passwords/password_controller.mm (right):

https://codereview.chromium.org/1950983002/diff/20001/ios/chrome/browser/pass...
ios/chrome/browser/passwords/password_controller.mm:244: web::WebState*
webState_;               // weak
On 2016/05/10 12:38:47, vasilii wrote:
> Why is the new member required?

Before it, the WebState was accessed through the WebStateObserverBridge. Caching
the WebState pointer passed to the initializer method avoids the need to create
the bridge before being able to vend the WebState.

[vasilii, 2016-05-10 12:52:00]

lgtm

[vabr (Chromium), 2016-05-10 12:54:35]

The CQ bit was checked by vabr@chromium.org

[commit-bot: I haz the power, 2016-05-10 12:54:58]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1950983002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1950983002/20001

[commit-bot: I haz the power, 2016-05-10 13:28:27]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-05-10 13:31:11]

Description was changed from

==========
PasswordController initialization: WebStateObserverBridge to the end

The associated bug has a crash where a WebStateObserverBridge method of
PasswordController dereferences a null pointer. The block where the dereference
happens is:

    if (weakSelf) {
      weakSelf.get()->passwordManager_->OnPasswordFormSubmitted(
          weakSelf.get()->passwordManagerDriver_.get(), form);
    }

While the most likely cause of that bug is being addressed in
https://codereview.chromium.org/1941363002/, the initialization order in
PasswordController's -initWithWebState is also suspicious: PasswordController
starts observing the WebState before it creates the supporting objects like
passwordManager_.


This CL changes the initialisation order to only create the bridge once the two
pointers are non-null.

BUG=608090
==========

to

==========
PasswordController initialization: WebStateObserverBridge to the end

The associated bug has a crash where a WebStateObserverBridge method of
PasswordController dereferences a null pointer. The block where the dereference
happens is:

    if (weakSelf) {
      weakSelf.get()->passwordManager_->OnPasswordFormSubmitted(
          weakSelf.get()->passwordManagerDriver_.get(), form);
    }

While the most likely cause of that bug is being addressed in
https://codereview.chromium.org/1941363002/, the initialization order in
PasswordController's -initWithWebState is also suspicious: PasswordController
starts observing the WebState before it creates the supporting objects like
passwordManager_.

This CL changes the initialisation order to only create the bridge once the two
pointers are non-null.

BUG=608090

Committed: https://crrev.com/bcd8636b075e7bcf9f5861f981695236dd533baf
Cr-Commit-Position: refs/heads/master@{#392588}
==========

[commit-bot: I haz the power, 2016-05-10 13:31:12]

Patchset 2 (id:??) landed as
https://crrev.com/bcd8636b075e7bcf9f5861f981695236dd533baf
Cr-Commit-Position: refs/heads/master@{#392588}