Fix handling of URLs with no schema.

Fix handling of URLs with no schema.

BUG=google-url:29
TEST=GURLTest::NonURL

[mnaganov (inactive), 2012-06-13 14:08:21]

Hi Brett and Will,

I have encountered this issue while working on Chrome for Android. Some of
Android CTS (compatibility) tests use completely invalid URLs (like the one I'm
using in the test case) just to check that the URL is passed correctly, with no
means of really using it. Such URLs get lost during IPC calls in Chrome because
URLs are serialized into strings using 'possibly_invalid_spec' and unmarshalled
by creating a new GURL instance from this string.

Can you please take a look at the proposed fix?

[willchan no longer on Chromium, 2012-06-13 21:37:05]

I think Brett's a better reviewer here. I'm removing myself. Cheers :)

[brettw, 2012-06-14 23:53:13]

lgtm

[mnaganov (inactive), 2012-06-15 08:26:37]

Thanks, Brett!

Will, can I kindly ask you to commit this change?

[willchan no longer on Chromium, 2012-06-15 17:18:08]

On 2012/06/15 08:26:37, Mikhail Naganov (Chromium) wrote:
> Thanks, Brett!
> 
> Will, can I kindly ask you to commit this change?

I'm going to try this fancy "commit" checkbox. Not sure if it works since this
is a different repo, but who knows?

[brettw, 2012-06-15 19:36:09]

It won't work.

[mnaganov (inactive), 2012-06-18 10:28:23]

On 2012/06/15 17:18:08, willchan wrote:
> On 2012/06/15 08:26:37, Mikhail Naganov (Chromium) wrote:
> > Thanks, Brett!
> > 
> > Will, can I kindly ask you to commit this change?
> 
> I'm going to try this fancy "commit" checkbox. Not sure if it works since this
> is a different repo, but who knows?

Will, can you please commit it in a regular way? Or if anyone could add me in
the project members list, I can commit it myself, I don't mind.

[willchan no longer on Chromium, 2012-06-18 20:18:39]

Sorry, I don't have an SVN client ready and I'm a bit overwhelmed today prepping
for a summit. Brett, can I punt committing this over to you?

On 2012/06/18 10:28:23, Mikhail Naganov (Chromium) wrote:
> On 2012/06/15 17:18:08, willchan wrote:
> > On 2012/06/15 08:26:37, Mikhail Naganov (Chromium) wrote:
> > > Thanks, Brett!
> > > 
> > > Will, can I kindly ask you to commit this change?
> > 
> > I'm going to try this fancy "commit" checkbox. Not sure if it works since
this
> > is a different repo, but who knows?
> 
> Will, can you please commit it in a regular way? Or if anyone could add me in
> the project members list, I can commit it myself, I don't mind.

[brettw, 2012-06-18 20:58:41]

I landed this as googleurl@176. I did not pull deps into Chrome. You should
submit a separate patch to update Chrome's DEPS to pull this revision.

[mnaganov (inactive), 2012-06-19 08:44:38]

On 2012/06/18 20:58:41, brettw wrote:
> I landed this as googleurl@176. I did not pull deps into Chrome. You should
> submit a separate patch to update Chrome's DEPS to pull this revision.

Awesome! Thank you very much, Brett! Of course I'll do the roll.

[brettw, 2012-08-20 19:02:04]

You did not do the roll, this breaks some Chrome tests, and now I have other
changes on top of this. I'm reverting this change.

[mnaganov (inactive), 2012-08-20 19:16:48]

On 2012/08/20 19:02:04, brettw wrote:
> You did not do the roll, this breaks some Chrome tests, and now I have other
> changes on top of this. I'm reverting this change.

Yes, I apologise. It turned out, that a lot of code implicitly relies on the
fact that GURL instances created with an invalid URL are in fact empty. I filed
a bug about that (CC'ing you) crbug/133509, and started fixing, but hadn't
finished, since this fix in GURL was only solving a small issue for us, and
spending so much time on fixing all affected code wasn't appropriate.

I'll try again later.