Integration Tests Http Bridge

[KtorZ]

Issue Number

#8

Overview

• I have added some integration tests which automatically test our API to fetch blocks from the http bridge
• I have removed unused code from the http bridge (according to the coverage report)
• I have added an invariant to the http bridge in case an invalid slot id is passed (I noticed that this will cause the algorithm to fetch all blocks from the starting epoch down to the initial one, which is quite bad).

Comments

[paweljakubas]

hmm, why it is all the time running? Because it was out-of-date at the time of pushing?

[KtorZ]

It seems that Travis failed to report the status to GitHub somehow. And, I realized I forgot to actually disable the integration tests for now in CI since we'll need a bit of caching first (or to run a local cluster)
😅

[paweljakubas]

I like the simplification adopted based on invariant and the plethora of tests added. The PR needs rebase and transformers only usage. Also it would be nice to be green in CI (I bet this is because it was not up-ti-date at the moment of push)

[paweljakubas]

this should be replaced with transformers

[paweljakubas]

Do we test (or it is reasonable to test) scenario when the tip refers to boundary block (ie., not main, but genesis). When talking with Byron nodes, one should take that into consideration. The consumers were aware of both types of blocks, handled two types of header hashes. But the blocks were dealt differently in two places of the code, in chain code (if I remember well) all blocks were downloaded and then filtered by consumer chain, some values form genesis block were used like time, ect.. Block coming from when listener mechanism (in the form of undos) were filtered upstream and the consumer (eg. wallet) were delivered only main blocks - it give rise to the need of that consumer to get additional data from the nodes (like time from slots, etc)

[KtorZ]

We do not because this is handled by our network layer. We do test fetching a whole epoch (so, when specifying a slot start at slot_number = 0). So, in contrary with the legacy wallet, we won't have to handle epoch boundaries blocks in the wallet logic (at this stage at least), since they are filtered out by the network layer.
When dealing with Haskell Shelley node, we'll have to reconsider some of that and will most probably have to apply some of the ledger rules to the EBB in order to keep track of a few protocol parameters.

[paweljakubas]

use rather import Control.Monad.Trans.Except

[akegalj]

@paweljakubas why is this so? Are we trying to use most explicit module in the import which exports the function?

[KtorZ]

cf: #81

[paweljakubas]

why it is called hermes? ;-)

[KtorZ]

No idea. Ask the rust team 😅