Curtail the occurrence of order books that are blocked by reduced offers
with the implementation of the fixReducedOffersV1 amendment.
This commit identifies three ways in which offers can be reduced:
1. A new offer can be partially crossed by existing offers, so the new
offer is reduced when placed in the ledger.
2. An in-ledger offer can be partially crossed by a new offer in a
transaction. So the in-ledger offer is reduced by the new offer.
3. An in-ledger offer may be under-funded. In this case the in-ledger
offer is scaled down to match the available funds.
Reduced offers can block order books if the effective quality of the
reduced offer is worse than the quality of the original offer (from the
perspective of the taker). It turns out that, for small values, the
quality of the reduced offer can be significantly affected by the
rounding mode used during scaling computations.
This commit adjusts some rounding modes so that the quality of a reduced
offer is always at least as good (from the taker's perspective) as the
original offer.
The amendment is titled fixReducedOffersV1 because additional ways of
producing reduced offers may come to light. Therefore, there may be a
future need for a V2 amendment.
* Use theoretical quality to order the strands
* Do not use strands below the user specified quality limit
* Stop exploring strands (at the current quality iteration) once any strand is non-dry
Historically strand re-execute log messages have been treated as
errors. However in the vast majority of cases these log messages
are caused by well understood mechanics in the payment engine.
So usually these log messages should be treated as warnings.
* In and Out parameters were swapped when calculating the rate
* In and out qualities were not calculated correctly; use existing functions
to get the qualities
* Added tests to check that theoretical quality matches actual computed quality
* Remove in/out parameter from qualityUpperBound
* Rename an overload of qualityUpperBound to adjustQualityWithFees
* Add fix amendment
The XRP Ledger utilizes an account model. Unlike systems based on a UTXO
model, XRP Ledger accounts are first-class objects. This design choice
allows the XRP Ledger to offer rich functionality, including the ability
to own objects (offers, escrows, checks, signer lists) as well as other
advanced features, such as key rotation and configurable multi-signing
without needing to change a destination address.
The trade-off is that accounts must be stored on ledger. The XRP Ledger
applies reserve requirements, in XRP, to protect the shared global ledger
from growing excessively large as the result of spam or malicious usage.
Prior to this commit, accounts had been permanent objects; once created,
they could never be deleted.
This commit introduces a new amendment "DeletableAccounts" which, if
enabled, will allow account objects to be deleted by executing the new
"AccountDelete" transaction. Any funds remaining in the account will
be transferred to an account specified in the deletion transaction.
The amendment changes the mechanics of account creation; previously
a new account would have an initial sequence number of 1. Accounts
created after the amendment will have an initial sequence number that
is equal to the ledger in which the account was created.
Accounts can only be deleted if they are not associated with any
obligations (like RippleStates, Escrows, or PayChannels) and if the
current ledger sequence number exceeds the account's sequence number
by at least 256 so that, if recreated, the account can be protected
from transaction replay.
This changes the rules for payments in two ways:
1) It sets the maximum number of offers any book step can consume from
2000 to 1000.
2) When a strand contains a step that consumes too many offers,
currently the liquidity is not used at all and the strand will
be considered dry. This changes things so the liquidity is used,
however the strand will still be considered dry.
* The compiler can provide many non-explicit constructors for
aggregate types. This is sometimes desired, but it can
happen accidentally, resulting in run-time errors.
* This commit assures that no types are aggregates unless existing
code is using aggregate initialization.
Replace Taker.cpp with calls to the payment flow() code.
This change required a number of tweaks in the payment flow code.
These tweaks are conditionalized on whether or not offer crossing
is taking place. The flag is explicitly passed as a parameter to
the flow code.
For testing, a class was added that identifies differences in the
contents of two PaymentSandboxes. That code may be reusable in
the future.
None of the Taker offer crossing code is removed. Both versions
of the code are co-resident to support an amendment cut-over.
The code that identifies differences between Taker and Flow offer
crossing is enabled by a feature. That makes it easy to enable
or disable difference logging by changing the config file. This
approach models what was done with the payment flow code. The
differencing code should never be enabled on a production server.
Extensive offer crossing unit tests are added to examine and
verify the behavior of corner cases. The tests are currently
configured to run against both Taker and Flow offer crossing.
This gives us confidence that most cases run identically and
some of the (few) differences in behavior are documented.
* Sanity check on newly created strands
* Better loop detection
* Better tests (test every combination of path element pairs)
* Disallow any root issuer (even for xrp)
* Disallow compount element typs in path
* Issue was not reset when currency was XRP
* Add amendment
If the mantissas of two non-native amounts differ by less than 10, then
subtracting them leaves a result of zero. This can cause situations
where `a>b`, yet `a-b == 0`.
One consequence of this is unfunded offers were incorrectly left in
order books. The code would check if the offer would be
consumed (`amount in offer > amount needed`), assume it wouldn't be,
yet when `amount needed` was subtracted from `amount in offer` the
result was zero and the offer was unfunded. This unfunded offer
incorrectly remained on the order book.
This patch fixes this bug.
The Ripple protocol represent transfer rates and trust line
qualities as fractions of one billion. For example, a transfer
rate of 1% is represented as 1010000000.
Previously, such rates where represented either as std::uint32_t
or std::uint64_t. Other, nominally related types, also used an
integral representation and could be unintentionally substituted.
The new Rate class addresses this by providing a simple, type
safe alternative which also helps make the code self-documenting
since arithmetic operations now can be clearly understood to
involve the scaling of an amount by a rate.
Before this change, the deferred credits algorithm took the current
balance and subtracted the recorded credits. Conceptually, this is the
same as taking the original balance, adding all the credits,
subtracting all the debits, and subtracting all the credits. The new
algorithm records the original balance and subtracts the debits. This
prevents errors that occur when the original balance and the recorded
credits have large differences in magnitude.
Additionally, XRP credits were recorded incorrectly in the deferred
credits table (the line was between the sender and receiver, rather than
the root account).
Payments do not remove unfunded and expired offers when a payment
fails. However, offer crossing is now using the payment engine and needs
to know what offers were found in a removable state, even on failure.
A computation like: `amount_remaining = amount_wanted - amount_got`, can
leave `amount_remaining == 0` without `amount_wanted == amount_got`.
This happens if the amounts differ by less than the smallest
representable value. Fix BookStep to handle this case.
Replace Journal public data members with member function accessors
in order to make Journal lighter weight. The change makes a
Journal cheaper to pass by value.
Also add missing stream checks (e.g., calls to JLOG) to avoid
text processing that ultimately will not be stored in the log.
Add a new algorithm for finding the liquidity in a payment path. There
is still a reverse and forward pass, but the forward pass starts at the
limiting step rather than the payment source. This insures the limiting
step is completely consumed rather than potentially leaving a 'dust'
amount in the forward pass.
Each step in a payment is either a book step, a direct step (account to
account step), or an xrp endpoint. Each step in the existing
implementation is a triple, where each element in the triple is either
an account of a book, for a total of eight step types.
Since accounts are considered in pairs, rather than triples, transfer
fees are handled differently. In V1 of payments, in the payment path
A -> gw ->B, if A redeems to gw, and gw issues to B, a transfer fee is
changed. In the new code, a transfer fee is changed even if A issues to
gw.
