digibyte-cli freezing during header sync

[saltedlolly]

Sometimes during startup digibyte-cli pauses for long stretches without returning a response - sometimes almost two minutes.

You can see a video showing the issue here. (I tried to post this on Github but it is too big) DigiNode Tools uses digibyte-cli throughout to check on the status of DigiByte Core. Notice that at 00:31 it pauses for almost two minues. This is the script simply waiting for digibyte-cli to respond. Not ideal. If it can't respond in a timely fashion it would be better to return an error rather than freezing.

While not a showstopping bug, perhaps this is something that can be optimized. It also may be something has has already been improved by #142 - more testing is needed. I am sharing this in case anyone else encounters the issue so we can look out for it. If all seems okay, we can close it.

@yacagel previously documented this in Issue #130 - #130

[JaredTate]

I think this is related to some of the memory leaks we have in the client. It's been a known issue we have had some memory leaks for years. I spent a few hours trying to track some of them down with macOS instruments & leaks tool along with lldb. I think we have some pointers in multialgo code we need to restructure somewhere that are not deallocating memory correctly. I will keep looking over the next couple weeks to see what I can dig up.

[JohnnyLawDGB]

From a non-technical perspective: I've experienced the header sync issue with digibyte-cli on occasion. I started looking at the network to try to understand what was occurring, as I appeared to be receiving inordinate amounts of data from nodes with high latency, without making progress. As I banned those specific nodes(temporarily) and the client connected to alternates, the synchronization began to make progress again.

[ghost]

Is this on the testnet?