It seems valid to pass in a null WebViewClient in the chromium port of WebKit, so check for that before invoking methods on it in the speech input code.
Created attachment 63050 [details] Patch
Comment on attachment 63050 [details] Patch r=me
Comment on attachment 63050 [details] Patch Clearing flags on attachment: 63050 Committed r64352: <http://trac.webkit.org/changeset/64352>
All reviewed patches have been landed. Closing bug.
FYI, it is more conventional to pass WebViewImpl* around within chromium/src/ than to pass WebViewClient*. You will frequently see code like this: if (viewImpl->client()) viewImpl->client()->someMethod()
Thanks for the comment, I'll use that pattern going forward and change this the next time this code is touched.