[cros] Add *Safe methods to the ownership API

[cros] Add *Safe methods to the ownership API

I was passing memory from libcros to chrome and expecting chrome to
free it.  I was also passing vectors across the DLL boundary.  These
could both cause memory horkage.  So, add safe methods and deprecate
the existing ones.

BUG=8283
TEST=unit tests, pull login_drive test binary over to device and test manually.

Change-Id: I7704caf3017947e9181c68c5efc627416abf561b

Committed: http://chrome-svn/viewvc/chromeos?view=rev&revision=8c0315a

[Chris Masone, 2010-10-28 19:53:11]

I hate that this change is so big, but it really seemed like the right thing to
do was add these structs and code that manipulates them...

[zel, 2010-10-28 20:02:46]

http://codereview.chromium.org/4132006/diff/14001/15003
File chromeos_login.cc (right):

http://codereview.chromium.org/4132006/diff/14001/15003#newcode322
chromeos_login.cc:322: // NOTREACHED() << "ChromeOSWhitelist is deprecated!";
you should just nuke these calls once chrome side is changed instead of having
NOTREACHED()

http://codereview.chromium.org/4132006/diff/14001/15005
File chromeos_login_helpers.cc (right):

http://codereview.chromium.org/4132006/diff/14001/15005#newcode40
chromeos_login_helpers.cc:40: uint8* result = static_cast<uint8*>(::operator
new(len));
what's that magic cast here for? what's wrong with simple 

uint8* result = new uint8[len];

[Chris Masone, 2010-10-28 20:09:08]

http://codereview.chromium.org/4132006/diff/14001/15003
File chromeos_login.cc (right):

http://codereview.chromium.org/4132006/diff/14001/15003#newcode322
chromeos_login.cc:322: // NOTREACHED() << "ChromeOSWhitelist is deprecated!";
On 2010/10/28 20:02:46, zel wrote:
> you should just nuke these calls once chrome side is changed instead of having
> NOTREACHED()

I figured we'd want to leave at least a bit of a window for older chrome
versions, no?

http://codereview.chromium.org/4132006/diff/14001/15005
File chromeos_login_helpers.cc (right):

http://codereview.chromium.org/4132006/diff/14001/15005#newcode40
chromeos_login_helpers.cc:40: uint8* result = static_cast<uint8*>(::operator
new(len));
On 2010/10/28 20:02:46, zel wrote:
> what's that magic cast here for? what's wrong with simple 
> 
> uint8* result = new uint8[len];

To stay in line with what's done in NewStringCopy in libchromeos.  I don't
understand exactly why sean did this incantation there, but I assumed that he
had a valid reason.

[zel, 2010-10-29 00:24:25]

LGTM