If GetNetworkList fails, don't push down an empty network list.

If GetNetworkList fails, don't push down an empty network list.

WebRTC would rather have a possibly-outdated network list than an
empty one. So if GetNetworkList fails, we now just log an error
instead of persisting the empty list.

BUG=616908
Committed: https://crrev.com/ba6d6a649dd1db1125dc6b321f1ad159aac0412c
Cr-Commit-Position: refs/heads/master@{#401021}

[Taylor_Brandstetter, 2016-06-17 22:36:22]

deadbeef@chromium.org changed reviewers:
+ sergeyu@chromium.org

[Sergey Ulanov, 2016-06-18 00:12:13]

Awesome, this looks like the same root issue that causes crbug.com/534463! (I
could never repro it, so didn't know the root cause).

I like the solution that Peter suggested on the bug:
when we get that error from GetAdaptersAddresses() try calling it again with the
new buffer size. WDYT?

That would fix the problem for both WebRTC and PPB_NetworkMonitor.
Also it probably makes sense to add a small delay between OnIPAddressChanged()
and DoGetNetworkList() to reduce probability of getting to that state.

[Sergey Ulanov, 2016-06-18 00:13:17]

On 2016/06/18 00:12:13, Sergey Ulanov wrote:
> Also it probably makes sense to add a small delay between OnIPAddressChanged()
> and DoGetNetworkList() to reduce probability of getting to that state.

The delay would also be useful to ensure that we get updated adapter list after
the OnIPAddressChanged() notification.

[Taylor_Brandstetter, 2016-06-20 22:45:26]

On 2016/06/18 00:12:13, Sergey Ulanov wrote:
> Awesome, this looks like the same root issue that causes crbug.com/534463! (I
> could never repro it, so didn't know the root cause).
> 
> I like the solution that Peter suggested on the bug:
> when we get that error from GetAdaptersAddresses() try calling it again with
the
> new buffer size. WDYT?
> 
> That would fix the problem for both WebRTC and PPB_NetworkMonitor.
> Also it probably makes sense to add a small delay between OnIPAddressChanged()
> and DoGetNetworkList() to reduce probability of getting to that state.

I have another CL to do just that: https://codereview.chromium.org/2071273004/

This CL is just in case there is some other issue we're not aware of that could
cause GetNetworkList to fail.

Also, I don't think the delay is necessary. After OnIPAddressChanged it's
guaranteed that the IP address that triggered the event is returned by
GetAdaptersAddresses. At least, that's what I inferred from the MSDN
documentation, and found in my testing.

[Sergey Ulanov, 2016-06-20 22:46:53]

lgtm

[Taylor_Brandstetter, 2016-06-21 16:11:48]

The CQ bit was checked by deadbeef@chromium.org

[commit-bot: I haz the power, 2016-06-21 16:12:17]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2078913004/1

[commit-bot: I haz the power, 2016-06-21 17:16:49]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2016-06-21 17:46:59]

Description was changed from

==========
If GetNetworkList fails, don't push down an empty network list.

WebRTC would rather have a possibly-outdated network list than an
empty one. So if GetNetworkList fails, we now just log an error
instead of persisting the empty list.

BUG=616908
==========

to

==========
If GetNetworkList fails, don't push down an empty network list.

WebRTC would rather have a possibly-outdated network list than an
empty one. So if GetNetworkList fails, we now just log an error
instead of persisting the empty list.

BUG=616908

Committed: https://crrev.com/ba6d6a649dd1db1125dc6b321f1ad159aac0412c
Cr-Commit-Position: refs/heads/master@{#401021}
==========

[commit-bot: I haz the power, 2016-06-21 17:47:00]

Patchset 1 (id:??) landed as
https://crrev.com/ba6d6a649dd1db1125dc6b321f1ad159aac0412c
Cr-Commit-Position: refs/heads/master@{#401021}