That explains it.
I don’t want to totally stop the wallet from updating the last known block. I could add a counter to have the wallet updated every 10 blocks. But changing the 2 sec parameter seems cleaner.
I ran some tests on a Raspberry Pi and indeed it fails to decode properly the downvotes. Until a fix is found the only option is to run it on another platform.
The costs also include many days off time of minting by many people and time spent in poking around before reporting. Generally we should stick to the KISS rule when more feature is requested.
Thans for the fix. But I won’t be able to use the Pi for a few weeks starting today so I can’t try the fix.
I have tried this on my pi with 2.0.3. The Pi has been sync’ing the blockchain for a stale client. Before I made the change the Pi flushed the wallets every 60 - 80 sec; each flush taking about 60-70 sec. That means the client has been flushing all the time almost. Every minutes only 2-3 blocks were sync’ed.
After I changed it to 200 seconds, te client flushes every 20-30 minunts. Sync;ing was much faster – 1 block / sec. But I felt that the flushint interval was too long. So I changed it to 60sec. Now it takes about 2-3 min to make one flush.
Actually since it only affects the ARM architecture which is not part of the released binaries, a 5.0.2 release would be exactly the same as the 5.0.1 one except for the version number.