You are viewing a single comment's thread from:
RE: A better approach to Turing Complete Smart Contracts
Releasing a pre-release of steem using memory mapped files today. Start up and shutdown times are now almost instant and with the rare exception of crashing in the middle of applying a block the chain database is almost never corrupted.
Is steemd able to identify that the database is corrupted on resume and do an automatic reindex, or do you just get undefined behavior? And same question but for the scenario of the computer suddenly shutting off in the middle of writing to the database?
If the computer shuts off in the middle of a write then it may result in undefined behavior; same if the program crashes while updating or if the file is modified by an outside program.
We plan to add a flag that we set when we start writing and we clear when we finish writing. Once this flag is in place then we can detect any crashes that leave the memory in undefined state.
The OS should take care of synchronizing the memory pages if the process crashes outside of a write operation.
The blockchain log is an append-only file now, which means it should never get corrupted and you should always be able to resync.