We got a report where a "tx already known" message caused a failed
trade but the deposit tx was valid.
To avoid such false positives we only handle reject messages which
we consider clearly critical.
Use a LinkedHashMap in place of a List, for the caching CurrencyUtil
fields 'allSortedFiatCurrencies' & 'allSortedCryptoCurrencies', using
the same iteration order as before. In this way, we can avoid a linear
search in the lookup methods getFiatCurrency & getCryptoCurrency.
In particular, this speeds up the activation of TradesChartsView (and to
a lesser extent OfferBookChartView), which make a lot of calls to
CurrencyUtil.getTradeCurrency in the fillTradeCurrencies/updateChartData
methods respectively.
At mediation we require a min. payout to the losing party to keep
incentive for the trader to accept the mediated payout. For Refund
agent cases we do not have that restriction.
According to https://github.com/bisq-network/proposals/issues/155
For buyer:
Default: 15%
Min. 15% for Altcoin, 15% for Fiat
Max. 50% for Altcoin, 50% for Fiat
Absolute min. deposit in BTC: 0.006 BTC (currently 40 USD)
For Seller:
Fixed: 15%
Absolute min. deposit in BTC: 0.006 BTC (currently 40 USD)
These are mostly injected objects that are now redundant, such as some
CoinFormatter and Preferences fields.
Also do some additional minor tidying of TradesChartsViewModel.
In case we had an unconfirmed change output we reset the
unconfirmedBsqChangeOutputList so that after a SPV resync we do not
have any dangling BSQ utxos in that list which would cause an incorrect
BSQ balance state after the SPV resync.
If a trader has a pending trade with an unconfirmed tx and is doing a
spv resync, the deposit tx is not found. They get displayed a popup
asking to restart and if problem continues to move trade to failed
trades. In regtest testing the missing deposit tx was always received
from the network at a restart. So this commit adds another button as
default button to shut down Bisq so that the user do first that activity
instead of moving the trade to failed trades.
It is not guaranteed that the trader will receive the missing deposit tx
at restart, but at least it is better as the state before. We should
find a way how to distinguish a valid unconfirmed tx from an invalid one
but it is not clear yet how we can do that and if it is feasible with
doing that with BitcoinJ only. It might require a external service to
look up the tx if it is in the mem pool, but even that will never gie a
100% certainty.
The internal 'java.io.DeleteOnExitHook' uses a LinkedHashSet of paths to
to delete files at when the JVM exits. However, since there is no way to
remove entries from the set, it leaks memory when calling deleteOnExit()
for every new temp file created by FileManager.saveToFile. To avoid
this, try to reuse file names (using one per FileManager instance).
(FileManager.saveToFile appears to be called on average around every 2s,
and this probably leaks around 36 + length * 2 bytes for the String
object, 36 bytes for the hashtable node and maybe around 10 bytes for
the table entry, or 82 + length * 2 ~= 236 bytes for a typical temp file
path length on Windows - about 10MB per day.)
As this private method is synchronized and not used in a way that
recursion/reentry is possible, any further instance-level locking within
the method is redundant - for each FileManager instance, there can only
be one saveToFile call in progress at a time.
* Remove unused parameters from assorted methods
Exclude abstract or default methods, as well as cases where the
parameter is currently unused but is probably intended to be used later.
* Actually use the injected Clock param of isDateInTolerance
Use 'clock.millis()' instead of "new Date().getTime()" in SignedWitness
& AccountAgeWitness, as the latter may have been left as an oversight.
Also tidy the date field of the toString() methods.
* Suppress warnings of unused method params which may be needed later
Also fix forwarding of telescoping method parameters in FormBuilder and
FormattingUtils.
Use 'clock.millis()' instead of "new Date().getTime()" in SignedWitness
& AccountAgeWitness, as the latter may have been left as an oversight.
Also tidy the date field of the toString() methods.
* Add user preference combobox for BSQ block explorer with random default
* Remove betanet and testnet BSQ block explorers
* Always check if a valid BSQ block explorer is set
