StopWallets, which was being called prior to UnloadWallets, performs an
unnecessary database closing step. This causes issues in UnloadWallets
which does additional database cleanups. Since the database closing step
is unnecessary, StopWallets is removed, and UnloadWallets is now called
from WalletLoaderImpl::stop.
chainStateFlushed is no longer needed since the best block is updated
after a block is scanned. Since the chainstate being flushed does not
necessarily coincide with the wallet having processed said block, it
does not entirely make sense for the wallet to be recording that block
as its best block, and this can cause race conditions where some blocks
are not processed. Thus, remove this notification.
Migrating a wallet requires having a bestblock record. This is always
the case in normal operation as such a record is always written on
wallet loading if it didn't already exist. However, within the unit
tests and benchmarks, this is not guaranteed. Since migration requires
the record, WalletMigration needs to also add this record before the
benchmark.
The only reason to call chainStateFlushed during wallet loading is to
ensure that the best block is written. Do these writes explicitly to
prepare for removing chainStateFlushed, while also ensuring that the
wallet's in memory state tracking is written to disk.
Additionally, after rescanning on wallet loading, instead of writing
the locator for the current chain tip, write the locator for the last
block that the rescan had scanned. This ensures that the stored best
block record matches the wallet's current state.
Any blocks dis/connected during the rescan are processed after the
rescan and the last block processed will be updated accordingly.
When a block is connected, if the new block had anything relevant to the
wallet, update the best block record on disk. If not, also sync the best
block record to disk every 144 blocks.
Also reuse the new WriteBestBlock method in BackupWallet.
a04f17a188 doc: warn that CheckBlock() underestimates sigops (Sjors Provoost)
Pull request description:
Counting sigops in the witness requires context that `CheckBlock()` does not have, so it only counts sigops for non-segwit transactions.
It's useful to document, but it should not be a problem.
The commit message contains some historical context.
ACKs for top commit:
ismaelsadeeq:
ACK a04f17a188
ryanofsky:
Code review ACK a04f17a188
Tree-SHA512: 26528367a7f3cfa8540ef0b90f7aa912c8f0bc057428f20a1fd1d4e232dac77747bc20044f0fcb0ffab8a2e1fb3dbe3dab46be749553a917744ddc7a829025cb
Made every signed/unsigned conversion in the serialization helpers explicit so the UBSan `implicit-sign-change` check passes and the `serialize.h` suppression can be dropped.
For consistency, a few other simple changes were also applied to the serialization helpers:
* remove redundant `inline` on function templates;
* unify formatting to make the differences between similar methods obvious.
- This commit creates a function `WaitTipChanged` that waits for the connected
tip to change until timeout elapsed.
- This function is now used by `waitTipChanged`
Co-authored-by: Ryan Ofsky <ryan@ofsky.org>
e7ad86e1ca test: fix another intermittent failure in wallet_basic.py (Martin Zumsande)
07350e204d test: Fix intermittent failure in wallet_basic.py (Martin Zumsande)
Pull request description:
Fixes two rare failures that happened in the CI:
#27249:
There could be a race with outstanding TxAddedToMempool notifications being applied to the soon-to-be created wallet:
1. importdescriptors during rescan sets status to `TxStateConfirmed`
2. old `transactionAddedToMempool` notification changes status back to `TxStateInMempool`
3. If the listunspent call happens here the test will fail
4. blockConnected notification will change the status back to `TxStateConfirmed` (so it's not a persistent failure)
I could reproduce this by adding a 100 microsecond sleep to `AddToWallet()`, the fix is to add a sync, so `transactionAddedToMempool` notifications won't affect the new wallet anymore.
#32456:
During init, the test framework will start using rpc after the mempool was loaded.
It will not wait for `start()` / `postInitProcess` or outstanding `transactionAddedToMempool` notifications (which would both set the status to `TxStateInMempool`), leading to
a possible race, in which `listunspent` can be called while the tx is still in `Inactive` status.
Can be reproduced by adding two sleeps: To init before calling `start()` for the chain clients, plus to `transactionAddedToMempool` in `wallet.cpp`.
Prevent this by processing outstanding notifications.
Fixes#27249Fixes#32456
ACKs for top commit:
maflcko:
review ACK e7ad86e1ca🎩
Tree-SHA512: 1f1a11e5c8e1c6d3c39a49401c2c5122befdbbec25c0451953f5bfe8dfb53221ada552a68006e266570addda12bb16c1b9b1e49ad2198c33d91c4b96b764d73e
fa427ffcee fuzz: Properly setup wallet in wallet_fees target (MarcoFalke)
Pull request description:
`g_wallet_ptr` is destructed after the `testing_setup`. This is not supported and will lead to issues such as https://github.com/bitcoin/bitcoin/pull/30221#issuecomment-2863875857 or https://github.com/bitcoin/bitcoin/pull/32409#issuecomment-2855259932.
This could be fixed by fixing the initialization order.
However, the global wallet is also modified in the fuzz target, which is bad fuzzing practise.
So instead fix it by constructing a fresh wallet for each fuzz iteration.
ACKs for top commit:
brunoerg:
code review ACK fa427ffcee
hebasto:
ACK fa427ffcee, this change fixes the issue when building the "Debug" configuration with MSVC on Windows.
marcofleon:
Code review ACK fa427ffcee
Tree-SHA512: 161b93fc39a609cb16d9ffea7366c5e339bd01712577f0782aedff46c00f79edd2a907807ac83f9fcec687b4bbbe0fd6e6f75e32169639a310e4e7b771078b3b
Enable it in the USDT tests. The context (from 0xB10C):
> every time the tracepoint is reached a SIGTRAP is fired.
> No matter the tracepoint contents, even with an empty one.
> Valgrind intercepts SIGTRAP and aborts.
See discussion in #32374.
It's probably debatable if we support targets like iOS, but for now,
document why we are checking for this standard library feature.
Trying to use `std::system` for a `aarch64-darwin-ios` target results in
```bash
test.cpp:7:10: error: 'system' is unavailable: not available on iOS
7 | std::system("some_command");
| ^
/Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/include/_stdlib.h:203:6: note: 'system' has been explicitly marked unavailable here
203 | int system(const char *) __DARWIN_ALIAS_C(system);
| ^
1 error generated.
```
rend() creates a pointer with offset -1. This is UB, according to the
C++ standard: https://eel.is/c++draft/expr.add#4:
When an expression J that has integral type is added to [...] an
expression P of pointer type, the result has the type of P.
... if P points to a (possibly-hypothetical) array element i of an
array object x with n elements [...] the expressions P + J and J + P
(where J has the value j) point to the (possibly-hypothetical) array
element i+j of x if 0≤i+j≤n [...]
Otherwise, the behavior is undefined.
Also, it is unclear why the functions exist at all, when stdlib utils
such as std::reverse_iterator{it} or std::views::reverse can be used out
of the box.
So remove them, along with the ubsan suppressions, that are no longer
used.
1e0de7a6ba fees: document non-monotonic estimation edge case (willcl-ark)
Pull request description:
Closes: https://github.com/bitcoin/bitcoin/issues/11800
In scenarios where data is available for higher targets but not for lower ones, this method *may* return lower fee rates for higher confirmation targets. This could occur if `estimateCombinedFee` returns no valid data (`-1`) for some estimates for a low target, but **does** return valid data for a higher target.
Users of this function should be aware of this potential, if unlikely, inconsistency in behaviour in data-sparse scenarios.
ACKs for top commit:
adamandrews1:
Code review ACK 1e0de7a
ismaelsadeeq:
Code review ACK 1e0de7a6ba
glozow:
ACK 1e0de7a6ba
Tree-SHA512: 161e5dafdd131570853a89491753ae39a7b725d1a86cab5a7294c2a5939da1a9a5f2c4aca0900e9ad810e828b6e0e636f256384e3d1fda6dd552da189bbbe747
0750249289 mining: document gbt_rule_value helper (Sjors Provoost)
5e87c3ec09 scripted-diff: rename gbt_force and gbt_force_name (Sjors Provoost)
Pull request description:
The term "force" is ambiguous and not used in [BIP9](https://github.com/bitcoin/bips/blob/master/bip-0009.mediawiki#getblocktemplate-changes) where there ! rule prefix is introduced.
E.g. this code is hard to read:
```cpp
if (!gbt_force) {
s.insert(s.begin(), '!');
```
Additionally, #29039 renamed `gbt_vb_name` to `gbt_force_name` which, at least for me, further increased the confusion.
This is a pure (variable rename) refactor (plus documentation) and does not change behavior.
Reminder of how to verify a scripted diff:
```sh
test/lint/commit-script-check.sh origin/master..HEAD
```
ACKs for top commit:
achow101:
ACK 0750249289
janb84:
ACK [0750249](0750249289)
musaHaruna:
ACK [0750249](0750249289)
glozow:
ACK 0750249289, seems sensible
Tree-SHA512: 8c88a273a3b36040f6c641843bd20579d0065b051aad4b39fc14f0d2af2808690dff6772bd8b1a4d9699b72279a700d2661012651bc315433a123dcc8996adaa
2aa63d511a test: Use uninvolved pruned node in feature_pruning undo test (enoch)
772ba7f9ce test: Fix nTimes typo in feature_pruning test (enoch)
Pull request description:
This PR contains two commits:
1. Fixes a typo in feature_pruning.py where 'nTimes' was incorrectly
used instead of 'nTime'. This typo caused the test to always reset
mine_large_blocks.nTime to 0, rather than only on the first run.
2. Fixes the test failure exposed by the typo fix. The
test_pruneheight_undo_presence test was failing because it was using
node 2, which is involved in reorg testing and could be on a
different chain than other nodes. The solution switches to using
node 5, which is also a pruned node but isn't involved in reorg
testing.
Testing:
- Ran test/functional/feature_pruning.py multiple times to verify
consistent passing
- Verified that the test now passes with the correct nTime variable name
- Confirmed the test behavior matches the intended functionality of
verifying pruned block availability
- Ran the full test suite to ensure the changes did not introduce any
regressions or affect other tests
Thanks to fjahr for his assistance in diagnosing the issue and
suggesting the solution.
This fixes the test failure reported in #32249
ACKs for top commit:
fjahr:
tACK 2aa63d511a
maflcko:
lgtm ACK 2aa63d511a
naiyoma:
tACK 2aa63d511a
stratospher:
tested ACK 2aa63d5. verified that `nTime` is being incremented now.
Tree-SHA512: a543528fd4eeb30e978c0b43cfa109768252edaf1f94679dbbc7fe684122c00da34224e2cc1abd2a265af1b267eef1cd34246207946cf7d8e93d2c0f11aa56d8
8673e8f019 txgraph: Special-case singletons in chunk index (optimization) (Pieter Wuille)
abdd9d35a3 txgraph: Skipping end of cluster has no impact (optimization) (Pieter Wuille)
604acc2c28 txgraph: Reuse discarded chunkindex entries (optimization) (Pieter Wuille)
c734081454 txgraph: Introduce TxGraph::GetWorstMainChunk (feature) (Pieter Wuille)
394dbe2142 txgraph: Introduce BlockBuilder interface (feature) (Pieter Wuille)
883df3648e txgraph: Generalize GetClusterRefs to support subsections (preparation) (Pieter Wuille)
c28a602e00 txgraph: Introduce TxGraphImpl observer tracking (preparation) (Pieter Wuille)
9095d8ac1c txgraph: Maintain chunk index (preparation) (Pieter Wuille)
87e74e1242 txgraph: abstract out transaction ordering (refactor) (Pieter Wuille)
2614fea17f txgraph: Add GetMainStagingDiagrams function (feature) (Pieter Wuille)
Pull request description:
Part of cluster mempool: #30289.
This adds more functionality to the txgraph module, specifically:
* `TxGraph::GetMainStagingDiagrams()`, a function to obtain feerate diagrams for both the main graph and the staged changes to it, including only the clusters that differ between the two.
* `TxGraph::GetBlockBuilder()`, a function to obtain an object which can efficiently iterate the chunks of the (main) graph from high to low chunk feerate, allowing each to be skipped or included.
* `TxGraph::GetWorstMainChunk()`, a function to obtain the last chunk that would be returned by `GetBlockBuilder()`'s returned object, intended for eviction.
ACKs for top commit:
monlovesmango:
reACK 8673e8f019
instagibbs:
reACK 8673e8f019
glozow:
reACK 8673e8f019
Tree-SHA512: 5c98c54919c44eb2f9545dfc130e54dfc25b5b54d43cf5ca9bcf46e019b9fd405a572fcd70e71e2a7c5b4b096cfd540a4d09ef1f52ba188504418682f1dfc4af
During init, the test framework will start using rpc after the
mempool was loaded. It will not wait for postInitProcess or
outstanding transactionAddedToMempool notifications, leading to
a possible race, in which listunspent is being called while the
tx is still in Inactive status. Prevent this by processing
outstanding notifications.
There could be a race with outstanding TxAddedToMempool notifications
being applied to the soon-to-be created wallet.
Fixes an intermittent timeout reproducable by adding a sleep to
AddToWallet.
- Create a new function `AddMerkleRootAndCoinbase` that compute the
block's merkle root, insert the coinbase transaction and the merkle
root into the block.
`-proxy=addr:port` specifies the proxy for all networks (except I2P).
Previously only the Tor proxy could have been specified separately via
`-onion=addr:port`.
Make it possible to specify separately the proxy for IPv4, IPv6, Tor and
CJDNS by e.g. `-proxy=addr:port=ipv6`. Or remove the proxy for a given
network, e.g. `-proxy=0=cjdns`.
Resolves: https://github.com/bitcoin/bitcoin/issues/24450
This interface lets one iterate efficiently over the chunks of the main
graph in a TxGraph, in the same order as CompareMainOrder. Each chunk
can be marked as "included" or "skipped" (and in the latter case,
dependent chunks will be skipped).
This is preparation for a next commit which will introduce a class whose
objects hold references to internals in TxGraphImpl, which disallows
modifications to the graph while such objects exist.
