Avoid mutating the Block tx list or the DaoState tx cache/index via a
Lombok getter. Instead wrap each in an unmodifiable[List|Map] & provide
specific mutator methods for use by DaoStateService to add newly parsed
transactions or load a DAO snapshot.
Also rename txMap to txCache, replace remaining use of getTxStream() in
the JSON file exporter with getUnorderedTxStream() (as this is safe) and
swap the arguments of the txCache initialisation merge function, for
exact consistency with the pre-caching behaviour.
Finally, add a missing assertDaoStateChange() and remove a potentially
harmful assertion from DaoStateService.onNewTxForLastBlock.
This is based on a suggested patch by @chimp1984 in the PR #3773 review.
Add getUnorderedTxStream() method to DaoStateService to stream directly
from the txMap cache/index wherever it is obviously safe to do so,
instead of iterating through the entire block list via getTxStream().
Also make getTxs() return a view of the txMap values in place of a copy.
This should improve efficiency slightly.
Build a HashMap of all BSQ transactions found, when loading the DaoState
from disc, and store it in a transient field which is always kept in
sync with the associated list of blocks. (The latter is only modified in
a couple of places in DaoStateService, making this straightforward.)
This is to speed up daoStateService.getTx(id), which is called from many
places and appears to be a significant bottleneck. In particular, the
initial load of the results in VoteResultView.doFillCycleList was very
slow (taking nearly a minute on a Core i3 machine) and likely to suffer
a quadratic slowdown (#cycles * #tx's) over time.
isDataOwner is used when deciding how many peer nodes should receive
a BroadcastMessage. If the BroadcastMessage originated
on the local node it is sent to ALL peer nodes with a small delay.
If the node is only relaying the message (it originated on a different
node) it is sent to MAX(peers.size(), 7) peers with a delay that is
twice as long.
All the information needed to determine whether or not the
BroadcastMessage originated on the local node is available at the final
broadcast site and there is no reason to have callers pass it in.
In the event that the sender address is not known during broadcast (which
is only a remote possibility due to how early the local node address
is set during startup) we can default to relay mode.
This first patch just removes the deep parameters. The next will remove
everything else. There is one real change in LiteNodeNetworkService.java
where it was using the local node when it should have been using the
peer node. This was updated to the correct behavior.
* Fix language tags so script and regional variants work correctly
* Adjust update_translations.sh language tags for script/regional variants
ACKs for top commit:
@ripcurlx:
ACK 089232716d
Fixes https://github.com/bisq-network/bisq/issues/3721
(part of the problem was that the trade ended up in failed trade)
Refactor method and add comments.
We did not handle the case of a mediated payout. isPayoutPublished() is
only reflecting non-disputed trade payouts.
Fixes https://github.com/bisq-network/bisq/issues/3721
(part of the problem was that the trade ended up in failed trade)
Refactor method and add comments.
We did not handle the case of a mediated payout. isPayoutPublished() is
only reflecting non-disputed trade payouts.
With v1.2 we use 2of2 multisig for deposit tx. This commit changes the
manual payout window to reflect that.
- Remove unused code from legacy arbitration
- Fix comments
With v1.2 we use 2of2 multisig for deposit tx. This commit changes the
manual payout window to reflect that.
- Remove unused code from legacy arbitration
- Fix comments
