GitLab

When decrypting responses to our XMLRPC requests, consider an older key to be OK. This fixes keyswap when both hosts have different keys (!)
(cherry picked from commit cfde5bfd)

Modify get_keypair so that it can be called from outside the class, and the regenerate parameter being true actually regenerates a key.

When an CryptException occurs when sending an XMLRPC request (because the response was using an old key for example), fail in a nicer manner.

Update the message for the primary cause of this (old key) so it's more obvious in the logs.

The MNET code uses this a bit.

Bump the default key life to 365 days, up from 28. We will be properly fixing keyswap soon, this is just a stopgap.
(cherry picked from commit f923e64c)

now find_remote_user returns user and authinstance that accepted them.

backed out change to user object to store the last xmlrpc auth instance.

along with all the other settings

- put back debug code I had inadvertently removed that didn't belong to
  me
- throw an exception i think is missing (someone please review this)
- better oo & namespacing
- better validation of incoming 'data' per format
- paths to zip and zip args (relies on a patch that i am about to put on
  master)

still to do:

- cron processing is completely untested
- i would like to split out the import table based on the transport:
  import_queue.host and token should go into a import_queue_mnet table
- it's possibly worth thinking about making import a proper plugin type.
  not sure about the effect this has on the import transport
  framework...  it might be possible to have both import and
  import transport plugintypes but that might be too heavy
- at the very least if we split out import_queue.host and token into an
  mnet table it will pave the way for a better refactor laterz.
- i would still really like to improve the mnet namespacing but that
  might be plausible at this point.
- need to write docs about arguments and return types
- i want to change the content_ready arguments to not include
  $filesmanifest as that is dependent on format being file - it may
  be actually better to dispatch somewhere else based on $format and
  then just have a generic $data which would be $filesmanifest for files
  and then something else for something like LEAP or maharanative or
  whatever, as this is checked in the importer, not not the
  importertransport.

The name is largely irrelevant as it's only the creator of the file.

Add a second parameter to get_peer, that will force re-retrieving a peer object rather than using a cached one.

Moodle's XMLRPC API currently makes some calls that Mahara doesn't implement. Implement these calls as no-ops.

Moodle uses it only in the course listing block, to say '4 courses from [othermoodle]'. Mahara doesn't have any courses.

Signed-off-by: Andrew McMillan

Now, table names (and other identifiers that could be confused with SQL kewords), should be put into SQL like this:

SELECT * FROM {artefact}

The braces are matched and expanded by DML now to include the prefix and be properly quoted, which means that tables like 'view' and 'group' don't need renaming (and nor should they, there's little reason why we should rename tables to get around such constraints).

This has removed a whole bunch of $prefix = ... and get_config('dbprefix') stuff, which makes things a little simpler yet again.

code

function

method to api_*

field need be passed to query