Add RegConnectRegistry to Advapi32 mappings

[cxcorp]

Adds the RegConnectRegistry function mapping to Advapi32. This function is used to connect to remote machines' registries.

I added a test case for the function, but to be completely fair, I'm not 100% sure if the \\localhost UNC path works. The test case could be edited to pass in null for the name, but then the test doesn't guarantee that the first parameter is a string.

[matthiasblaesing]

Thank you - in general this looks good. Two things I'd like to ask before a merge:

• was this code tests in a real setting with a remote attach? If not, could you please do it?
• the unittest looks sane - I assume it fails, if the remote registry service is not started - so if there is a separate error code, that indicates that, it might be worth checking for, if the error is unspecific, the test should be marked manual (See the SAFEARRAY tests, there is one I marked to be run only manually)

[cxcorp]

Hiya! Thanks for taking a look at the PR.

I just tested this on my localhost with the Remote Registry service started. It works as intended:

A user on StackOverflow also claimed (in a comment) that the mapping works.

As for the errors, it seems to return ERROR_BAD_NETPATH (0x35) if the remote service is disabled or if the path is just garbage. Doesn't seem to differentiate between the service being disabled and being unable to connect to the host. If attempting to use access a HKEY which is not one of the three supported ones, it returns ERROR_INVALID_HANDLE (0x6). I don't have an environment to test workgroup misconfigurations, lacking privileges or invalid credentials in, so I don't know which errors they might return. However, a Symantec support page I found suggests (in my opinion) that RegConnectRegistry may also return ACCESS_DENIED (0x5) if the user lacks privileges to the registry on the machine.

Should I make the test check for ERROR_BAD_NETPATH and leave a comment that it assumes remote registry is disabled?

[matthiasblaesing]

Please adjust the unittest to check for ERROR_BAD_NETPATH and output a describing message to System.err, this way an interested party can see the result. This should not cause the unittest to fail.

Please also update CHANGES.md in the toplevel directory. You should add an entry to the features section for 5.0.0 describing your contribution. Have a look at the over entries and follow that pattern.

[matthiasblaesing]

@cxcorp can I help you with the necessary changes?

[cxcorp]

Hey, I've been busy during the weekdays but I'll get around to these changes this weekend!

…

On Tue, Mar 27, 2018, 21:08 Matthias Bläsing ***@***.***> wrote: @cxcorp <https://github.com/cxcorp> can I help you with the necessary changes? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#935 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AUIjFOa0eXF1fR3RkvnBi5XAse4NnGKVks5tioA2gaJpZM4SciX5> .

[matthiasblaesing]

Hey Joona, Am Mittwoch, den 28.03.2018, 05:17 -0700 schrieb Joona Heikkilä:

I've been busy during the weekdays but I'll get around to these changes this weekend!

absolutely ok - it just wanted to make sure all is well. If you are busy I can totally understand, so no worries. Greetings Matthias

[cxcorp]

Hiya, I made changes to the test, docs and CHANGES.md. Could you take a look?

[matthiasblaesing]

Thank you - looks good and is merged to master.