Previously, for endpoints bound to agents, we would pass the request
onto the agent even if the agents wasn't currently running.
Here, we make eyre check to see if the agent is actually running, before
passing the request on. If the bound agent is not running, eyre serves a
503 synchronously instead.
This way, we avoid cluttering up the gall queue for the bound agent.
For the server case, we add tests for http-first finalization, incorrect
eauth tokens, attempt expiration, client abort signalling, and client
deletion requests.
For the client case, we add a test for approving eauth attempts without
being locally authenticated.
To this end, we refactor the layout of the eauth tests. The +eauth core
now contains the tasks/gifts to be handled by the client/server, but no
longer does move checking. This happens within the test arms only. This
way, we may trivially re-run parts of the eauth flow under different
conditions.
Concatenating before we truncate, instead of truncating the entropy by
itself, is slightly simpler.
Because this slightly changes the naming algorithm, we must update the
eyre tests to match.
This shaves off 1000 lines of testing code while maintaining the same
tests. It reduces boilerplate by introducing "mare", a monad for
testing Eyre. It's very simple (just maintains the current state of
Eyre and the current time), but it's easier to build helper functions in
this form, and that reduces the immense quantities of copy-and-paste
that were in the old tests. What's there now could surely be improved
further, but I think this is a good start.
The underlying mare machinery is not really specific to Eyre, so it
would be straightforward to apply this strategy to other vanes. The
work is in creating appropriate helper functions for each vane. Eyre is
undergoing work, so that's the only one I've changed here. Further,
it's not clear that this is the ultimate solution to unit testing vanes.
The resulting code is IMO clearer than before, but I wouldn't say it's
*clear*.
Unauthenticated requests now also create sessions. This affects most
HTTP request handling tests.
The situation here is not ideal, and worsening over time. Worth spending
some time to think about how to best refactor the Eyre tests to make
them more manageable and easier to maintain.
|pump and |sink call into each other in three places
related to nacks and naxplanations (sending a nack,
notifying the |pump of a naxplanation, or dropping a
nack from the |sink). This intra calls are making implicit
updates to more parts of the state than the core should
manage. To avoid that we emit a move to %arvo, encoded
as an %ames plea, to handle that in the next event.
Since the tests call arms from eyre directly, these had to be updated
alongside it, to now specify the channel mode in light of which the
request should be parsed.
Includes smoke tests for jam-mode channels.
previously we were acking the nack-trace message (adding a %send move)
before notifying the message-pump with a %near task. Now, due to the
refactoring we invert the order of those moves. This seems safe but will
determine with livenet testing
Including these in pkg/arvo has caused many minor problems over the
years. We don't want to include them in releases, but this often meant
excluding them manually, which was error-prone.
Here, we move them out of /pkg/arvo/tests and into /tests. CI will have
to be updated to match, since we'll still want to include tests there.
