Reduce buggy usage of the registry controlled domain service.

Reduce buggy usage of the registry controlled domain service.

GetRegistryLength for host names canonicalizes the input for the caller, but
then returns the length in the canonicalized input, which is not necessarily
the same as the length in the original string. As a result, computations
performed by the caller based on this value can be wrong (see the bug for
more).

All callers of this function were audited and changed to use on of the
following:

- Many callers don't need the offsets. A new function
  HostHasRegistryControlledDomain is added to check for the presence of
  a R.C.D. without the risk of returning incorrect string lengths.

- Many callers already have guaranteed-canonical strings (they came out of
  a GURL or KURL object soon before the call) These were changed to use a
  new GetCanonicalHostRegistryLength function. A further advantage is that
  these calls will be faster.

- A new Permissive function is added that handles cases where the input
  is necessarily non-canonical.

Adds an IDN test case to the unit tests.

Removes checking for IP addresses in the already-known-canonical cases.
This requires a separate full canonicalization and IP addresses should
never match the R.C.D. list.

Reland of https://codereview.chromium.org/2446273004/ with fix which was a
reland of https://codereview.chromium.org/2433583002/ with fix.

TBR=pkasting@chromium.org
BUG=657199
Committed: https://crrev.com/1b0d95acde33777b9b9ab136c35eeefc11a64718
Cr-Commit-Position: refs/heads/master@{#427908}

[brettw, 2016-10-26 21:57:08]

Fix

[brettw, 2016-10-26 21:57:09]

The CQ bit was checked by brettw@chromium.org to run a CQ dry run

[brettw, 2016-10-26 21:57:47]

I submitted the last patch when the android_cronet builder (non-default) went
green, but that apparently isn't everything :( Will wait for all cronet try runs
this time.

[commit-bot: I haz the power, 2016-10-26 21:58:08]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[brettw, 2016-10-26 22:13:39]

The CQ bit was unchecked by brettw@chromium.org

[brettw, 2016-10-26 22:49:34]

The CQ bit was checked by brettw@chromium.org

[commit-bot: I haz the power, 2016-10-26 22:50:11]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-10-27 01:44:16]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2016-10-27 01:46:02]

Description was changed from

==========
Reduce buggy usage of the registry controlled domain service.

GetRegistryLength for host names canonicalizes the input for the caller, but
then returns the length in the canonicalized input, which is not necessarily
the same as the length in the original string. As a result, computations
performed by the caller based on this value can be wrong (see the bug for
more).

All callers of this function were audited and changed to use on of the
following:

- Many callers don't need the offsets. A new function
  HostHasRegistryControlledDomain is added to check for the presence of
  a R.C.D. without the risk of returning incorrect string lengths.

- Many callers already have guaranteed-canonical strings (they came out of
  a GURL or KURL object soon before the call) These were changed to use a
  new GetCanonicalHostRegistryLength function. A further advantage is that
  these calls will be faster.

- A new Permissive function is added that handles cases where the input
  is necessarily non-canonical.

Adds an IDN test case to the unit tests.

Removes checking for IP addresses in the already-known-canonical cases.
This requires a separate full canonicalization and IP addresses should
never match the R.C.D. list.

Reland of https://codereview.chromium.org/2446273004/ with fix which was a
reland of https://codereview.chromium.org/2433583002/ with fix.

TBR=pkasting@chromium.org
BUG=657199
==========

to

==========
Reduce buggy usage of the registry controlled domain service.

GetRegistryLength for host names canonicalizes the input for the caller, but
then returns the length in the canonicalized input, which is not necessarily
the same as the length in the original string. As a result, computations
performed by the caller based on this value can be wrong (see the bug for
more).

All callers of this function were audited and changed to use on of the
following:

- Many callers don't need the offsets. A new function
  HostHasRegistryControlledDomain is added to check for the presence of
  a R.C.D. without the risk of returning incorrect string lengths.

- Many callers already have guaranteed-canonical strings (they came out of
  a GURL or KURL object soon before the call) These were changed to use a
  new GetCanonicalHostRegistryLength function. A further advantage is that
  these calls will be faster.

- A new Permissive function is added that handles cases where the input
  is necessarily non-canonical.

Adds an IDN test case to the unit tests.

Removes checking for IP addresses in the already-known-canonical cases.
This requires a separate full canonicalization and IP addresses should
never match the R.C.D. list.

Reland of https://codereview.chromium.org/2446273004/ with fix which was a
reland of https://codereview.chromium.org/2433583002/ with fix.

TBR=pkasting@chromium.org
BUG=657199

Committed: https://crrev.com/1b0d95acde33777b9b9ab136c35eeefc11a64718
Cr-Commit-Position: refs/heads/master@{#427908}
==========

[commit-bot: I haz the power, 2016-10-27 01:46:04]

Patchset 2 (id:??) landed as
https://crrev.com/1b0d95acde33777b9b9ab136c35eeefc11a64718
Cr-Commit-Position: refs/heads/master@{#427908}

[foolip, 2016-10-27 09:58:38]

A revert of this CL (patchset #2 id:20001) has been created in
https://codereview.chromium.org/2459493002/ by foolip@chromium.org.

The reason for reverting is: Failing DCHECK_EQ in GetCanonicalHostRegistryLength
on WebKit Mac10.11 (dbg) running
http/tests/xmlhttprequest/origin-whitelisting-ip-addresses-with-subdomains.html

BUG=659949.