Remove unsafe registry APIs from base::win::RegKey

Remove unsafe registry APIs from base::win::RegKey

Clean up any callers hitting new DCHECKs.

BUG=375400

[Will Harris, 2017-06-28 15:48:25]

Description was changed from

==========
Remove unsafe registry APIs from base::win::RegKey

Clean up any callers hitting new DCHECKs.

BUG=375400
==========

to

==========
Remove unsafe registry APIs from base::win::RegKey

Clean up any callers hitting new DCHECKs.

BUG=375400
==========

[Will Harris, 2017-06-28 15:48:25]

wfh@chromium.org changed reviewers:
+ grt@chromium.org

[Will Harris, 2017-06-28 15:50:14]

WIP

this doesn't work yet. Because remoting is using RegKey::Set and RegKey::Take()
all over the place

https://cs.chromium.org/search/?q=%22Take(%22+file:%5Esrc/remoting/host/+pack...

offering it up in this unfinished state to see if you have any ideas, in
particular is this even worth doing, or should I give up... I think I should be
able to do Set and Take with move semantics... but then there's that Duplicate
Handle in pairing_registry_delegate_win.cc - I suppose I could implement that
using copy semantics?

WDYT?

[grt (UTC plus 2), 2017-06-29 08:49:47]

i've been wanting RegKey to be a movable type for some time. if you have the
time/patience for it, i think it would be awesome to make RegKey more like File.
i think this would be a big enough undertaking that it should be planned and
executed in stages. what do you think, do you have time for it? if not, some
small changes to make it more safe are welcome.

as for pairing_registry_delegate_win, would it help to introduce:

// Returns a new instance with a distinct handle to the same key as |this|.
RegKey RegKey::Duplicate() {
}

if Duplicate were to take a REGSAM, it could use RegOpenKeyEx rather than
DuplicateHandle. i'm not sure if one way is better than the other for any
reason.

fwiw, looks like me2me_native_messaging_host_main.cc is leaking handles when it
calls SetRootKeys. is that what you're trying to fix once and for all with this
change?

https://codereview.chromium.org/2963783002/diff/20001/base/win/registry.cc
File base/win/registry.cc (right):

https://codereview.chromium.org/2963783002/diff/20001/base/win/registry.cc#ne...
base/win/registry.cc:158: DCHECK(key_);
i'm not sure about this. when you don't care about error handling, it's nice to
be able to open a key and try to operate on it. if the open fails, so does
everything else. does RegCreateKeyEx return a reasonable result if key_ is 0?

https://codereview.chromium.org/2963783002/diff/20001/chrome/common/chrome_pa...
File chrome/common/chrome_paths.cc (right):

https://codereview.chromium.org/2963783002/diff/20001/chrome/common/chrome_pa...
chrome/common/chrome_paths.cc:127: if
(FAILED(path_key.ReadValue(kFlashPlayerPathValueName, &path_str)))
ugh, FAILED is wrong