Allow features to refer to subkeys in _manifest_features.json.

Allow features to refer to subkeys in _manifest_features.json.

Manifest restriction violations are now warnings instead of fatal errors.

Add background.persistent subkey and restrict it to dev mode.

BUG=122459
TEST=no

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=134596

[Matt Perry, 2012-04-25 01:06:26]

http://codereview.chromium.org/10217017/diff/1/chrome/common/extensions/manif...
File chrome/common/extensions/manifest.cc (right):

http://codereview.chromium.org/10217017/diff/1/chrome/common/extensions/manif...
chrome/common/extensions/manifest.cc:160: return true;
This is technically a change in functionality (needed because not all subkeys
are features), but I think it shouldn't matter in practice, because we never
call Get() for non-features.

[Matt Perry, 2012-04-25 01:22:41]

Hmm.. I changed ValidateManifest to emit warnings instead of errors, but on
second thought, I think we still want some things to be fatal. Will have to look
more at this.

[Matt Perry, 2012-04-25 20:50:17]

OK, ready for another look. ValidateManifest now gives fatal errors instead of
warnings again. I think we'll have to change this on a case-by-case basis.

I also removed some of the tests that verified that you can't query unknown keys
from the manifest. I changed that behavior because it doesn't seem useful.

[Aaron Boodman, 2012-04-25 21:48:31]

http://codereview.chromium.org/10217017/diff/5001/chrome/common/extensions/ap...
File chrome/common/extensions/api/_manifest_features.json (right):

http://codereview.chromium.org/10217017/diff/5001/chrome/common/extensions/ap...
chrome/common/extensions/api/_manifest_features.json:19: "extension",
"packaged_app", "hosted_app", "platform_app"
Omit 'hosted_app'.

http://codereview.chromium.org/10217017/diff/5001/chrome/common/extensions/ma...
File chrome/common/extensions/manifest_unittest.cc (left):

http://codereview.chromium.org/10217017/diff/5001/chrome/common/extensions/ma...
chrome/common/extensions/manifest_unittest.cc:135: // Verifies that the various
getters filter unknown and restricted keys.
Are there any checks of restricted keys here? It looks like they are all just
unknown.

You didn't change the behavior of restricted keys, and that still seems useful
and should be tested.

[Matt Perry, 2012-04-25 22:28:18]

OK, ValidateManifest is back to emitting non-fatal warnings.

http://codereview.chromium.org/10217017/diff/5001/chrome/common/extensions/ap...
File chrome/common/extensions/api/_manifest_features.json (right):

http://codereview.chromium.org/10217017/diff/5001/chrome/common/extensions/ap...
chrome/common/extensions/api/_manifest_features.json:19: "extension",
"packaged_app", "hosted_app", "platform_app"
On 2012/04/25 21:48:31, Aaron Boodman wrote:
> Omit 'hosted_app'.

Done.

http://codereview.chromium.org/10217017/diff/5001/chrome/common/extensions/ma...
File chrome/common/extensions/manifest_unittest.cc (left):

http://codereview.chromium.org/10217017/diff/5001/chrome/common/extensions/ma...
chrome/common/extensions/manifest_unittest.cc:135: // Verifies that the various
getters filter unknown and restricted keys.
On 2012/04/25 21:48:31, Aaron Boodman wrote:
> Are there any checks of restricted keys here? It looks like they are all just
> unknown.
> 
> You didn't change the behavior of restricted keys, and that still seems useful
> and should be tested.

They were all for unknown keys. I added a new test for restricted keys.

[Matt Perry, 2012-04-26 01:13:36]

Hmm... This change allows someone to create a theme with extension elements.
We'll install it, but give a warning. Do we want that?

[Aaron Boodman, 2012-04-26 01:27:56]

On Wed, Apr 25, 2012 at 6:13 PM,  <mpcomplete@chromium.org> wrote:
> Hmm... This change allows someone to create a theme with extension elements.
> We'll install it, but give a warning. Do we want that?

The extension elements should be ignored. If so, that seems OK to me.

- a

[Matt Perry, 2012-04-27 20:25:40]

OK, I updated the InstallTheme test to take into account the change in behavior.

[Aaron Boodman, 2012-04-28 02:27:48]

LGTM