The checkpoints were formerly copied from the full node, so based
on possible fork-events.
That seems less useful nowadays and since we want to be able to
allow starting a blockchain instance from any checkpoint it makes
sense to spread them out more evenly.
Based on the concept that the BCH chain generates around 50K blocks
a year, we simply use that interval.
This removes various very old ones meaning that the total count is
roughly equal.
This improves the docs, adds a getter and last it makes the
blockHeightAtTime() more secure by not returning a value below the
checkpoint. As this would throw if used directly in the block() method.
This public method allows an application to replace the static-
header files in a life-situation. Without needing to restart the app.
The idea is that a static chain starts at a checkpoint, and we extend
it into the past to start at an earlier checkpoint. This suddenly
makes available a larger amount of history without changing anything
of the data that was already there.
The blockchain object holds all block-headers, the backing store for
this class is now being exposed more to allow understanding what we have
or not have available. And also what the source of that info is.
As we moved most of the creation of a BufferPool to be via the
Streaming::pool() method, which uses a thread-local, it makes sense
to start cleaning up the design and make it more modern C++.
The above mentioned method would return a reference and you'd see
loads of places use `auto &pool =` which is less than ideal.
As the number of places where we actually instantiate a BufferPool
goes down, the usage of some sort of smart pointer makes more sense.
This now makes all APIs use BufferPool be wrapped in a shared_ptr.
The system writes a new file every single run (provided new headers were
received) and we sometimes compact them into a big file again.
The code forgot to remove the newly introduced info files of the old
files it compacted. Leaving confusing things happening after.
This solves that by making the first run remove all info files and re-
build them, adding a version byte to allow us freedom to do that in the
future again.
We optimistically create a new info file but as we start we might
instantly realize the file is useless and give up before having written
a single byte.
We now remove that file to avoid stale state.
Also be more verbose on warnings.
The blockchain class can effectively now be shallow,
we can have a list of blockheaders of, for instance, the last year
alone. It builds on top of a known checkpoint (hardcoded block data)
and as long as we do not access block info that is unavailable, things
just work like normal.
We throw at the request of a blockheader that is too old.
The methodname:
blockHeightAtTime()
makes the reader think about the actual height of the chain at that
time. Returning that value until the timestamp of the next block is
reached.
Now the method actally returns that number.
The code does not do any work to account for blocks going backwards in
time compared to the block before.
Makes the change from yesterday complete: we now
compress all the chunks into one periodically.
This, again, is done with a timeout instead of on-close.
We have 45 seconds as a timeout to balance between need to save and
waiting long enough to get additional headers should it
have been a while since last run.
This change makes it so that we rely less on the downloaded
blockheader data being only in memory and us saving it when
the client shuts down.
Instead, we save more regularly and we save smaller files
instead of always overwriting the complete downloaded headers.
Having less to save at shutdown is then a happy side-effect.
What we did previously is load a static chain which was to avoid lots of
work by simply memory-mapping the data instead of inserting it in a
vector.
This improves upon that by also having a metadata file which holds the
block-hashes directly readable. This avoids us having to iterate over
the blockheaders and hashing them.
Additionally we now have a single chain-work field that is the total
amount added work for the headers. This saves us from doing 256-bit
divisions and work for each blockheader as well.
Result is that we make the statup mostly about memory moving and remove
all CPU intensive stuff. Going from 1450ms to 400ms on desktop.
I expect an even greater gain on mobile CPUs.
A peer that is behind is marked as "on a different chain", which is
ultimately correct, but the response to punish or ban them is too much.
We could very well reconnect later when the situation is resolved.
So, detect when a peer is simply behind and respond by disconnecting
without punishment.
Instead of assigning a privacy segment at connect, wait until we have
determined if this peer is on the same chain as us before we allocate a
peer slot.
This also helps us limit the amount of peers we send our bloom filter to.
This is done in several steps:
1. Separate my height from the remote peer heights.
Instead of assuming we have one height, recognize that a peer may
not be at the tip at the same time we are. We monitor headers/invs
to update the 'peerHeight' variable.
2. Ask for merkle blocks from a peer to the maximum height of that
peer (but not later than what we validated to be true).
This avoids us asking past the remotes tip which they didn't like.
3. Redo the SyncSPVAction to use all this and make it much more
reliable in finding peers to download from and getting all the
changes as fast as possible.
tomFlowee
Your Name