[Mac] Chromium builds shouldn't ask for access to Chrome's own secure data

[Mac] Chromium builds shouldn't ask for access to Chrome's own secure data.

The message that this presents is:

Chromium wants to use your confidential information stored in “Chrome Safe
Storage” in your keychain. Do you want to allow access to this item?

Chromium shouldn't ask for access to Google Chrome's data, it should have its
own Chromium Safe Storage item in the keychain.

R=isherman@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=261515

[Mark Mentovai, 2014-04-01 19:16:25]

->isherman

This is different from the address book prompt you’re working on, but several of
us agreed that Chromium shouldn’t be messing with Chrome’s data in this way.

[Ilya Sherman, 2014-04-02 01:32:29]

Hmm, after this change, will Chromium builds be able to read passwords from
Safari but not from Chrome?  Or will they be able to read passwords from both,
but just not overwrite ones generated by Chrome?  IMO it would be very weird to
end up in a state where we can read Safari's passwords, but not Chrome's.

[Mark Mentovai, 2014-04-02 16:32:31]

The Keychain item in question isn’t a password, it’s a key used to encrypt some
data stored in the profile. Since Chrome and Chromium have their own distinct
profiles, it’s not a problem for them to have distinct keys. See
components/os_crypt/os_crypt.h. Users of this API include sync and autofill, to
avoid writing things like credit card numbers to disk in plaintext.

[Ilya Sherman, 2014-04-02 22:48:46]

On 2014/04/02 16:32:31, Mark Mentovai wrote:
> The Keychain item in question isn’t a password, it’s a key used to encrypt
some
> data stored in the profile. Since Chrome and Chromium have their own distinct
> profiles, it’s not a problem for them to have distinct keys. See
> components/os_crypt/os_crypt.h. Users of this API include sync and autofill,
to
> avoid writing things like credit card numbers to disk in plaintext.

I see.  Makes sense, LGTM.

[Mark Mentovai, 2014-04-03 20:37:25]

Committed patchset #1 manually as r261515.