[chromedriver] Add JavaScript fallback for GoBack/GoForward if DevTools commands aren't available

[chromedriver] Add JavaScript fallback for GoBack/GoForward if DevTools commands aren't available

BUG=chromedriver:980
Committed: https://crrev.com/64015fe05756a754b30bb7a6c24a2191b5765d4f
Cr-Commit-Position: refs/heads/master@{#307102}

[samuong, 2014-12-05 02:13:41]

samuong@chromium.org changed reviewers:
+ stgao@chromium.org

[samuong, 2014-12-05 02:13:42]

I've tested this on my phone (which runs Lollipop) and the DevTools commands
work, but for older versions of WebView (e.g. for the KitKat device on our
waterfall) we still need to use the JavaScript method.

We shouldn't run into the problem of missed responses because the older versions
of WebView won't ignore the old RenderViewHost messages.

[stgao, 2014-12-05 21:59:18]

lgtm with nits.

https://codereview.chromium.org/783593003/diff/1/chrome/test/chromedriver/chr...
File chrome/test/chromedriver/chrome/web_view_impl.cc (right):

https://codereview.chromium.org/783593003/diff/1/chrome/test/chromedriver/chr...
chrome/test/chromedriver/chrome/web_view_impl.cc:181: if (message ==
"'Page.getNavigationHistory' wasn't found")
Is it possible to use browser info like build_number or blink_revision for this
backward-compatibility check?
The error message might change and seems not that reliable, and I remember the
error code is always the same for all errors from DevTools.

If not, a TODO and a bug for cleaning this up might help.

https://codereview.chromium.org/783593003/diff/1/chrome/test/chromedriver/chr...
chrome/test/chromedriver/chrome/web_view_impl.cc:218: return
Status(kUnknownCommand);
Maybe use kUnknownError instead.
Chrome layer should know nothing about command.

[samuong, 2014-12-05 22:12:11]

https://codereview.chromium.org/783593003/diff/1/chrome/test/chromedriver/chr...
File chrome/test/chromedriver/chrome/web_view_impl.cc (right):

https://codereview.chromium.org/783593003/diff/1/chrome/test/chromedriver/chr...
chrome/test/chromedriver/chrome/web_view_impl.cc:181: if (message ==
"'Page.getNavigationHistory' wasn't found")
On 2014/12/05 21:59:18, Shuotao wrote:
> Is it possible to use browser info like build_number or blink_revision for
this
> backward-compatibility check?
> The error message might change and seems not that reliable, and I remember the
> error code is always the same for all errors from DevTools.
> 
> If not, a TODO and a bug for cleaning this up might help.

Unfortunately not, WebView reports its build number as 0. I'll add that to the
comment as well as a TODO.

https://codereview.chromium.org/783593003/diff/1/chrome/test/chromedriver/chr...
chrome/test/chromedriver/chrome/web_view_impl.cc:181: if (message ==
"'Page.getNavigationHistory' wasn't found")
On 2014/12/05 21:59:18, Shuotao wrote:
> Is it possible to use browser info like build_number or blink_revision for
this
> backward-compatibility check?
> The error message might change and seems not that reliable, and I remember the
> error code is always the same for all errors from DevTools.
> 
> If not, a TODO and a bug for cleaning this up might help.

Done.

https://codereview.chromium.org/783593003/diff/1/chrome/test/chromedriver/chr...
chrome/test/chromedriver/chrome/web_view_impl.cc:218: return
Status(kUnknownCommand);
On 2014/12/05 21:59:18, Shuotao wrote:
> Maybe use kUnknownError instead.
> Chrome layer should know nothing about command.

Done.

[samuong, 2014-12-05 22:12:22]

The CQ bit was checked by samuong@chromium.org

[commit-bot: I haz the power, 2014-12-05 22:13:01]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/783593003/20001

[commit-bot: I haz the power, 2014-12-05 23:09:27]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2014-12-05 23:10:18]

Patchset 2 (id:??) landed as
https://crrev.com/64015fe05756a754b30bb7a6c24a2191b5765d4f
Cr-Commit-Position: refs/heads/master@{#307102}