422 | @@ -423,7 +423,7 @@ std::string HelpMessage(HelpMessageMode mode)
423 | strUsage += HelpMessageOpt("-limitdescendantsize=<n>", strprintf("Do not accept transactions if any ancestor would have more than <n> kilobytes of in-mempool descendants (default: %u).", DEFAULT_DESCENDANT_SIZE_LIMIT));
424 | strUsage += HelpMessageOpt("-bip9params=deployment:start:end", "Use given start/end times for specified BIP9 deployment (regtest-only)");
425 | }
426 | - string debugCategories = "addrman, alert, bench, coindb, db, http, libevent, lock, mempool, mempoolrej, net, proxy, prune, rand, reindex, rpc, selectcoins, tor, zmq"; // Don't translate these and qt below
427 | + string debugCategories = "addrman, alert, bench, block{,2}, coindb, db, estimatefee{,2}, http, libevent, lock, mempool{,rej,2}, net{,2}, proxy, prune, rand, reindex, rpc, selectcoins, tx{,2}, tor, zmq"; // Don't translate these and qt below
I do not like shellishms like block{,2}.
what other options are there? I think this is going to be a situation where it'll be impossible to please everyone but I'm happy to modify it to go with the majority.
I think that for debugging, one category is enough...
@paveljanik Thanks for the feedback. I can certainly merge block2 and block together, if this is what most people want. In an older version I had historical block upload logged, and this generated a lot of output, so block2 was created mostly for this. Currently that logging is disabled entirely which makes block2 less necessary, so perhaps it can be gone already. I'll review the code now, but I think yes block2 debug can be disabled - although I do prefer to give people options, hence leaving it in. If people don't want it, they don't need to enable it.
In an earlier version I would show timings, i.e. how long the delay between requesting headers/blocks and them being received was - this was useful to refining timeout code and identifying when download was stalled to a far better degree than the current code. i.e. it would determine within 10 seconds with high reliably when a node had stalled rather than the current 2 minutes. (it did this by monitoring the partial download and actively keeping track of typical delay times).
@paveljanik Thinking about your suggestion further, I think block does need to have at least two subsets, as the logging of block invs received for old blocks would be too verbose for most people, and yet I suspect some people would want to enable this. (after all, it was enabled with "net" before this pull). People might also ask for the debug of block uploads, but since this wasn't in the code before this pull I felt it was reasonable to leave it still out (and find a better way to log it more concisely for a future pull request).
