Use a new function in BitcoinTestFramework (get_node_args) to specify which options each node should have.
I guess this is a matter of taste, but in my opinion this makes the code cleaner (also for future changes to the options). It would even allow to specify "general" options that are set differently for the four nodes (e. g., two could always have -txindex and two not, so that both behaviours can be tested).
Use a new function in BitcoinTestFramework (get_node_args) to specify
which options each node should have.
I can't see the benefit of this PR. The way of settings args (in current master) is appropriate IMO. Remind also that this PR adds ~150 lines of code.
Therefore i tend to NACK.
I think that it makes the argument clearer. Also, it could be used to get completely rid of the setup_network functions in some of the tests if taken a bit further. I wanted to submit this first to get initial feedback before spending more time, though. But yes, I realise that it is probably a matter of taste.
Note that at least some of the added lines of code are probably due to more prevalent line breaks in my coding style than in some of the replaced code. Also more savings will come when removing some setup_network functions becomes possible.
I like test cases to be nitty-gritty and explicit. This makes them easier to understand, or at least, makes it easier to understand what actually happens. Deeper abstractions make it harder to troubleshoot, or for 'outsiders' to change the tests when necessary for code changes.
So I also tend towards NACK here. Currently we specify the exact arguments that are fed into bitcoind in every test. This would bury it in more Python syntactic sugar.