The migration benchmark crashes if run more than once, because of `std::move(wallet)` and leaves subsequent iterations in an undefined state - avoiding `UndefinedBehaviorSanitizer` null‑dereference error.
`MigrateLegacyToDescriptor` returns both a spendable descriptor wallet and a watch‑only wallet.
If these remain attached, their files stay open and on Windows this can hang CI when removing the test directory.
By constructing them via `MakeWalletLoader` (which owns the `WalletContext`), both wallets are automatically unloaded when the loader is destroyed at the end.
This ensures no lingering handles or resource leaks when running the benchmark on CI with `-sanity-check`.
Co-authored-by: furszy <matiasfurszyfer@protonmail.com>
fa86190e6e rpc: Allow fullrbf fee bump (MarcoFalke)
Pull request description:
The RPCs (psbt)bumpfee, and the GUI, reject fee bumps when BIP 125 signalling is absent in the transaction even when the mempool and other RPCs allow them. Fix the confusion by allowing the fee bump.
This is done after fullrbf is always on (https://github.com/bitcoin/bitcoin/pull/30592)
ACKs for top commit:
1440000bytes:
reACK fa86190e6e
achow101:
ACK fa86190e6e
w0xlt:
ACK fa86190e6e
rkrux:
reACK fa86190e6e
glozow:
ACK fa86190e6e
Tree-SHA512: b2ffe8dcadbe71e9be767a16cf8aa0bf383c2de7aa1aee9438d125f444e24f3f7e4f02ddb28981bd3b8b645b6a24a407b4ad6bb0b21946ae637e78f6386e05bf
acee5c59e6 descriptors: Have GetPrivKey fill keys directly (Ava Chow)
4b0303197e descriptors: Move FlatSigningProvider pubkey filling to GetPubKey (Ava Chow)
25a3b9b0f5 descriptors: Have GetPubKey fill origins directly (Ava Chow)
6268bde0af descriptor: Remove unused parent_info from BIP32PUbKeyProvider::GetPubKey (Ava Chow)
0ff072caa1 wallet, rpc: Only allow keypool import from single key descriptors (Ava Chow)
Pull request description:
Instead of having `MakeScripts` infer what pubkeys need to go into the output `FlatSigningProvider`, have each of the `PubkeyProviders` that have `GetPubKey` and `GetPrivKey` called fill it directly with relevant keys and origins.
This allows for keys and origins to be added that won't directly appear in the output, which is necessary for `musig()` descriptors.
Split from #29675
ACKs for top commit:
fjahr:
Code review ACK acee5c59e6
theStack:
re-ACK acee5c59e6
rkrux:
ACK acee5c5
Tree-SHA512: c1841359bcb08cdd433122deef96579236928660785f3357a3eb584e47d290cd1c60ebe8f7fba50f178ba45c9a90773124e0f509e36c5a0df97c1a4890e03e5c
e3d7533ac9 test: improves tapscript unit tests (Ethan Heilman)
3e167085ba test: Ensures test fails if witness is not hex (Ethan Heilman)
Pull request description:
This commit creates new test utilities for future Taproot script tests within script_tests.json. The key features of this commit are the addition of three new tags: `#SCRIPT#`, `#CONTROLBLOCK#`, and `#TAPROOTOUTPUT#`. These tags streamline the test creation process by eliminating the need to manually generate these components outside the test suite.
* `#SCRIPT#`: Parses Tapscript and outputs a byte string of opcodes.
* `#CONTROLBLOCK#`: Automatically generates the control block for a given Taproot output.
* `#TAPROOTOUTPUT#`: Generates the final Taproot scriptPubKey.
This code was originally part of the OP_CAT PR https://github.com/bitcoin/bitcoin/pull/29247 but was pulled out into a separate PR to reduce the rebase treadmill for the OP_CAT PR.
Additionally this PR adds a check to ensure that if the witness data can not be parsed as hex the test fails. Prior to this PR, the test code would fail silently and set the values it couldn't parse as empty stack elements. This fix was suggested by @instagibbs.
## Rationale
While writing JSON script tests (script_tests.json) for https://github.com/bitcoin/bitcoin/pull/29247 we ran into the following problem. The JSON script tests are simple and easy to write for pre-Tapscript scripts, but adding or changing a Tapscript test requires substantial work per test. Consider the following pre-tapscript test:
```
["'aa' 'bb'", "CAT 0x4c 0x02 0xaabb EQUAL", "P2SH,STRICTENC", "DISABLED_OPCODE", "CAT disabled"]
````
whereas a Tapscript test for the same script (annotated with comments for better readability) would look like:
```
[
[
"aa",
"bb",
"7e4c02aabb87", // output script
"c0d6889cb081036e0faefa3a35157ad71086b123b2b144b649798b494c300a961d", // control block
0.00000001
],
"",
"0x51 0x20 0x15048ed3a65748549c27b671936987093cf73a4c9cb18522a74fb9553060ca99", // Tapscript output
"P2SH,WITNESS,TAPROOT",
"OK",
"TAPSCRIPT CATs aa and bb together and checks if EQUAL to aabb"
]
```
Computing the Tapscript output, such as `0x51 0x20 0x15048ed3a65748549c27b671936987093cf73a4c9cb18522a74fb9553060ca99`, requires writing custom code and running it for each test. The same is true for the Tapscript control block, such as `c0d6889cb081036e0faefa3a35157ad71086b123b2b144b649798b494c300a961d`. If a test is changed or updated new outputs and control blocks must be computed. The complexity of doing this is likely the reason that no one has added any Tapscript tests to JSON script tests until this PR.
In this PR we address this issue by adding the following improvements to JSON script tests:
Adding simple macros ("#SCRIPT# and #CONTROLBLOCK#) that allow the script test parser to automatically generate and inject a valid Tapscript output and control block to be computed automatically from the JSON script.
Allowing Tapscript scripts to use the human readable strings like pre-script scripts by marking the location of the script in the witness stack using #SCRIPT#. This transforms the unreadable script 7e4c02aabb87 into #SCRIPT# CAT 0x4c 0x02 0xaabb EQUAL.
This results in the following JSON script test which is far easier to write and easier to read.
```
[
[
"aa",
"bb",
"#SCRIPT# CAT",
"#CONTROLBLOCK#",
0.00000001
],
"",
"0x51 0x20 #TAPROOTOUTPUT#",
"P2SH,WITNESS,TAPROOT,OP_CAT",
"OK",
"TAPSCRIPT Test of OP_CAT flag by calling CAT on two elements. TAPSCRIPT_OP_CAT flag is set so CAT is executed."
],
```
ACKs for top commit:
instagibbs:
reACK e3d7533ac9
sipa:
utACK e3d7533ac9
janb84:
Re ACK [e3d7533](e3d7533ac9)
Tree-SHA512: 948c3ec28a4b2b222c2d77e48918ed19d298b51d64662fc20959073edd9978fc796516a392da9755a7e173f556e3021816dc6ce8eb3ed16bbe0fa6ebc574fd48
This commit creates new test utilities for future Taproot script
tests within script_tests.json. The key features of this commit are the
addition of three new tags: `#SCRIPT#`, `#CONTROLBLOCK#`, and
`#TAPROOTOUTPUT#`. These tags streamline the test creation process by
eliminating the need to manually generate these components outside the
test suite.
* `#SCRIPT#`: Parses Tapscript and outputs a byte string of opcodes.
* `#CONTROLBLOCK#`: Automatically generates the control block for a given
Taproot output.
* `#TAPROOTOUTPUT#`: Generates the final Taproot scriptPubKey.
Update src/test/script_tests.cpp
Co-authored-by: Jan B <608446+janb84@users.noreply.github.com>
18a035145d ci: Temporarily disable `WalletMigration` benchmark (Hennadii Stepanov)
Pull request description:
The `WalletMigration` benchmark is currently failing on CI.
This PR temporarily disables it until the issue is resolved.
An alternative to https://github.com/bitcoin/bitcoin/pull/32302.
ACKs for top commit:
maflcko:
lgtm ACK 18a035145d
TheCharlatan:
ACK 18a035145d
Tree-SHA512: bb1451fd0743a2955216a6d06916e411420a76bfed8b69ffcfadf99d0996d8f3b89ed72f855f25269f943ca4c3b4422065fde2374a1bf76c8bb64f14ab883092
d85895e5a7 build, msvc: Build only required `qtbase` features (Hennadii Stepanov)
fe5a6dcc53 build, msvc: Update vcpkg manifest baseline (Hennadii Stepanov)
Pull request description:
This PR:
1. Updates the vcpkg manifest baseline from the [2024.09.30 Release](https://github.com/microsoft/vcpkg/releases/tag/2024.09.30) to the [2025.03.19 Release](https://github.com/microsoft/vcpkg/releases/tag/2025.03.19), with the following package changes:
- boost: 1.85.0#1,2 --> 1.87.0
- qtbase: 6.7.2#3 -> 6.8.2#1
- qttools: 6.7.2#1 -> 6.8.2
- sqlite3: 3.46.1 --> 3.49.1
The previous update was made in https://github.com/bitcoin/bitcoin/pull/31186.
3. Explicitly specifies required features for the `qtbase` package, which makes vcpkg skip unused features such as `dnslookup`, `openssl`, etc.
ACKs for top commit:
hodlinator:
ACK d85895e5a7
Tree-SHA512: e05fb50825c6ba2e0caf8ded1dfb55fa42f28f06dcd28597f748c1843a1c4cc888e62ad87b6481c66481a576fa224dd4943fe76e286ff579c6d4dc2ec6a78a71
5cb1241814 feefrac: avoid integer overflow in temporary (Pieter Wuille)
Pull request description:
In `FeeFrac::Div(__int128 n, int32_t d, bool round_down)` in src/util/feefrac.h, the following line computes the result:
```c++
return quot + (mod > 0) - (mod && round_down);
```
The function can only be called under conditions where the result is in range, and thus doesn't involve any integer overflow. However, the intermediary result computed by just `quot + (mod > 0)` may still overflow if it's going to be corrected by the `- (mod && round_down)` that follows.
Fix this by balancing the two correction steps with each other first:
```c++
return quot + ((mod > 0) - (mod && round_down));
```
Fixes#32294.
ACKs for top commit:
l0rinc:
Tested ACK 5cb1241814
maflcko:
lgtm ACK 5cb1241814
achow101:
ACK 5cb1241814
Tree-SHA512: 9daaccdf9acd7652d53b52cad2dc12872558265e863acdde2d6015f885cb87c0505f9bd5be5499fc0a0eded29bec719643f6af1fbc3604518143985094226c95
e261eb8d50 tests: Add BIP 373 test vectors (Ava Chow)
26370c68d0 rpc: Include MuSig2 fields in decodepsbt (Ava Chow)
ff3d460898 psbt: Implement un/ser of musig2 fields (Ava Chow)
Pull request description:
Implements un/serialization of MuSig2 PSBT fields and prepares PSBT to be able to sign for MuSig2 inputs.
Split from #29675
ACKs for top commit:
fjahr:
re-ACK e261eb8d50
theStack:
re-ACK e261eb8d50
rkrux:
tACK e261eb8d50
Tree-SHA512: bb852ad074978847ac4dc656332025e2d4d1025d4283537b89618c7cadd61a8ecd2eff24779b8a014bc8d7b431125060449768192fa05ad0577f29e3c64b2374
2835216ec0 txgraph: make GroupClusters use partition numbers directly (optimization) (Pieter Wuille)
c72c8d5d45 txgraph: compare sequence numbers instead of Cluster* (bugfix) (Pieter Wuille)
Pull request description:
Part of cluster mempool: #30289
The implicit transaction ordering for transactions in a TxGraphImpl is defined by:
1. higher chunk feerate first
2. lower Cluster* object pointer first
3. lower position within cluster linearization first.
Number (2) is not deterministic, as it intricately depends on the heap allocation algorithm. Fix this by giving each Cluster a unique `uint64_t m_sequence` value, and sorting by those instead.
The second commit then uses this new approach to optimize GroupClusters a bit more, avoiding some repeated checks and dereferences, by making a local copy of the involved sequence numbers.
Thanks to @dergoegge for pointing this out.
ACKs for top commit:
instagibbs:
reACK 2835216ec0
marcofleon:
ACK 2835216ec0
glozow:
utACK 2835216ec0
Tree-SHA512: d772a55b9ed620159b934a42a39fca7f900d4aa89c099a280a0c61ea0bd7c4fc39b388281ffc775064ea77b0b17263871b4c9763aa71c710a79287d5eb2cd4b4
fa6a007b8e fuzz: Avoid integer sanitizer warnings in policy_estimator target (MarcoFalke)
Pull request description:
It seems odd to write a fuzz target to trigger integer sanitizer warnings in `CBlockPolicyEstimator::processBlockTx` and then suppress them. If the scenario can happen in reality, the code should be properly fixed to handle the cases. If not, it seems better to fix the fuzz target to not trigger meaningless traces.
Do that here by keeping track of the current height and limiting mempool entries to at most this entry height.
ACKs for top commit:
brunoerg:
ACK fa6a007b8e
dergoegge:
utACK fa6a007b8e
Tree-SHA512: 2092017dc309fb095fe5d43cfb76efb691795f303d567ee919be2b5cac19a944293636229903dc4d1e8b9fe5daf9dc3058544321eff1735f91f804c3baa36cd0
Also, fix the incorrect documention of the 'replaceable' RPC argument
with respect to sequence number handling. The docs were incorrect
before, so the fix could be extracted, but it seems fine to include here
as well.
9ccee9cd02 doc: Document WITH_EXTERNAL_LIBMULTIPROCESS build option better (Ryan Ofsky)
Pull request description:
This includes a cmake documentation change suggested https://github.com/bitcoin/bitcoin/pull/31741#discussion_r2039716492 and another change to mention the option in markdown documentation
ACKs for top commit:
hebasto:
ACK 9ccee9cd02, changes look good.
TheCharlatan:
ACK 9ccee9cd02
Tree-SHA512: c9103b001b970ac57afedc6dc384091f5661975d569573e93003cbd7df1891c54cefb06d7296eac5b9a5c57251803dcab2bd3b26c9d81aa476c62f211dcb3d6e
27f11217ca ci: drop -priority-level from bench in win cross CI (fanquake)
Pull request description:
So there's at least one CI sanity checking all benchmarks.
Related to #32277.
ACKs for top commit:
l0rinc:
utACK 27f11217ca
hebasto:
ACK 27f11217ca.
mabu44:
utACK 27f11217ca
Tree-SHA512: 4853584bf9db418f6e31aa0f558d08bc45479d672b193e1d25a25907f82fb225bc4388321f8f23286cd9fd9168c7546c713829607eb0cf5e3c62b98e88f8e68b
faeb1babe2 ci: refactor: Use version id over version codename consistently (MarcoFalke)
fae322a43a ci: Slim down lint image (MarcoFalke)
3333273a8f ci: Bump lint imagefile FROM base (MarcoFalke)
Pull request description:
Currently, the lint_test_runner is built and installed into the lint CI image. This is problematic, because it triggers a full image build on every change to its source code. Doing a build of the lint test_runner on every run is easier and faster.
ACKs for top commit:
l0rinc:
ACK faeb1babe2
janb84:
Re- ACK [faeb1ba](faeb1babe2)
Tree-SHA512: 39103e61ec2587096213bc1ce55b80087f6f03775592827d8c96a366453b798570d912690bf96fde4685798e5fc8ee2695ce851f473b4c8782d1a4c50c65a594
7912cd4125 bench: Fix WalletMigration benchmark (pablomartin4btc)
Pull request description:
The keys and scripts created for the Legacy Wallet needed to be persisted in order for the migration to work properly.
Fixes#32277.
ACKs for top commit:
achow101:
ACK 7912cd4125
davidgumberg:
Tested ACK 7912cd4125
furszy:
utACK 7912cd4125
Tree-SHA512: fe7b8e0a80d4d030ad3fd6446717ee09a260ab2bd6140bc817bdca52d233e3af8a8fed2d754743ca2ba022f7d2c8615a36b5070991d12942c13835e8f72e359f
7a93544cdc doc: Fix and clarify description of ZMQ message format (Jiri Jakes)
Pull request description:
This change stresses that all ZMQ messages share the same structure and that they differ only in the format of the bodies. Previously this was not clear.
Further it removes the notion of endianness of 32-byte hashes, as it was misleading, and replaces it with the term 'reversed byte order' (as opposed to natural or normal byte order produced by hashing functions).
Additionally, it states that ZMQ 32-byte hashes are in the same format as in RPC. Previously it incorrectly stated that the two were in different formats.
[Rendered](https://github.com/jirijakes/bitcoin/blob/zmq-doc/doc/zmq.md).
Fixes#31856.
ACKs for top commit:
w0xlt:
Code review ACK 7a93544cdc
achow101:
ACK 7a93544cdc
ryanofsky:
Code review ACK 7a93544cdc. Nice changes. Documentation seems less repetitive and easier to understand now
Tree-SHA512: 8c5ab047c5fd9b5b6910d691b725886d7743dfd01510735b46e43d01c2d0d25ec52d79d71ec75dbeb142e96a88ad503d69ee14b971e3cdaeb8fd85e5292a8c21
a4041c77f0 test: Handle empty string returned by CLI as None in RPC tests (Brandon Odiwuor)
Pull request description:
Partially Fixes https://github.com/bitcoin/bitcoin/issues/32264
Some tests are failing when `bitcoin-cli` returns an empty string. This change treats an empty response as `None`. See https://github.com/bitcoin/bitcoin/issues/32264#issuecomment-2807616694
This fixes the error for:
- feature_bip68_sequence.py
- feature_nulldummy.py
- feature_signet.py
- mining_mainnet.py
- rpc_scanblocks.py
- rpc_scantxoutset.py
- wallet_descriptor.py --descriptors
ACKs for top commit:
maflcko:
lgtm ACK a4041c77f0
achow101:
ACK a4041c77f0
pablomartin4btc:
ACK a4041c77f0
mzumsande:
ACK a4041c77f0
Tree-SHA512: 2f1a416a18e0b3eebdb014c2e2e8dadf1d46b15c231cb61f577d47f5e551994ab0e2aeb7c179c01be7c1f07ebc03476236d29cf2d04c358ffb1fae985aa385c9
32dcec269b rpc: update RPC help of `createpsbt` (rkrux)
931117a46f rpc: update the doc for `data` field in `outputs` argument (rkrux)
8134a6b5d4 rpc: add cli example for `walletcreatefundedpsbt` RPC (rkrux)
Pull request description:
### add cli example for `walletcreatefundedpsbt` and `createpsbt` RPCs
The only example present earlier was one that creates an OP_RETURN output. This
lack of examples has discouraged me earlier to use this RPC. Adding an example
that creates PSBT sending bitcoin to address, a scenario that is much more common.
### rpc: update the doc for `data` field in `outputs` argument
It was not evident to me that this field creates an `OP_RETURN` output until
I read the code and tried it out. Thus, making the doc explicitly mention it.
This affects docs of the following RPCs:
`bumpfee`, `psbtbumpfee`, `send`, `walletcreatefundedpsbt`, `createpsbt`,
and `createrawtransaction`
ACKs for top commit:
sipa:
utACK 32dcec269b
1440000bytes:
utACK 32dcec269b
achow101:
ACK 32dcec269b
ryanofsky:
Concept ACK 32dcec269b. These seem like helpful clarifications, but I did not look into the details
Tree-SHA512: f994488ba7d52d00960fc52064bb419cf548e29822fe23d6ee0452fdf514dd93f089145eddb32b8086a7918cf8cf33a4c3f16bfcb7948f3c9d5afd95e8d3a1cb
7749d929a0 Remove support for RNDR/RNDRRS for aarch64 on Linux (laanwj)
Pull request description:
This hardware feature is
- Rarely supported on SoCs (and broken on like half of the chips that support it in the first place) (#31817). It is not clear if, or how, the brokenness will be worked around in the kernel, but working around it in user space seems the wrong thing to do, this is not the place to maintain special workarounds for specific hardware (which despite that, was attempted in #31826, but had to be reverted in #31908 due to other problems).
- Apparently not compiled into the release binary anymore (https://github.com/bitcoin/bitcoin/issues/31817#issuecomment-2795885962). Did check this at the time, but a build system change must have caused this, and went undetected.
- Hard to test in CI (as well as manually), due to unavailability of hardware.
Better to remove it.
This reverts commit aee5404e02 from #26839.
Closes#31817.
ACKs for top commit:
sipa:
utACK 7749d929a0
davidgumberg:
utACK 7749d929a0
achow101:
ACK 7749d929a0
w0xlt:
utACK 7749d929a0
Tree-SHA512: d243ad7f745fb46f711f24b6983d9ea1d94e5d8ee60959229bafdba5caa210a60801a1c2cb5b558a0e72f365371b32285aee9a8d0cd24a60589adc7b03dd6a44
Instead of GetPrivKey returning a key and having the caller fill the
FlatSigningProvider, have GetPrivKey take the FlatSigningProvider and
fill it by itself. This will be necessary for descriptors such as
musig() where there are private keys that need to be added to the
FlatSigningProvider but do not directly appear in any resulting scripts.
GetPrivKey is now changed to void as the caller no longer cares whether
it succeeds or fails.
Instead of having ExpandHelper fill in the origins in the
FlatSigningProvider output, have GetPubKey do it by itself. This reduces
the extra variables needed in order to track and set origins in
ExpandHelper.
Also changes GetPubKey to return a std::optional<CPubKey> rather than
using a bool and output parameters.
Legacy wallets should only import keys to the keypool if they came in a
single key descriptor. Instead of relying on assumptions about the
descriptor based on how many pubkeys show up after expanding the
descriptor, explicitly mark descriptors as being single key type and use
that for the check.
05117e6e17 rpc: clarify longpoll behavior (Sjors Provoost)
5315278e7c Have createNewBlock() wait for a tip (Sjors Provoost)
64a2795fd4 rpc: handle shutdown during long poll and wait methods (Sjors Provoost)
a3bf43343f rpc: drop unneeded IsRPCRunning() guards (Sjors Provoost)
f9cf8bd0ab Handle negative timeout for waitTipChanged() (Sjors Provoost)
Pull request description:
This PR prevents Mining interface methods from sometimes crashing when called during startup before a tip is connected. It also makes other improvements like making more RPC methods usable from the GUI. Specifically this PR:
- Adds an `Assume` check to disallow passing negative timeout values to `Mining::waitTipChanged`
- Makes `waitfornewblock`, `waitforblock` and `waitforblockheight` RPC methods usable from the GUI when `-server=1` is not set.
- Changes `Mining::waitTipChanged` to return `optional<BlockRef>` instead of `BlockRef` and return `nullopt` instead of crashing if there is a timeout or if the node is shut down before a tip is connected.
- Changes `Mining::waitTipChanged` to not time out before a tip is connected, so it is convenient and safe to call during startup, and only returns `nullopt` on early shutdowns.
- Changes `Mining::createNewBlock` to block and wait for a tip to be connected if it is called on startup instead of crashing. Also documents that it will return null on early shutdowns.
This allows `waitNext()` (added in https://github.com/bitcoin/bitcoin/pull/31283) to safely assume `TipBlock()` isn't `null`, not even during a scenario of early shutdown.
Finally this PR clarifies long poll behaviour, mostly by adding code comments, but also through an early `break`.
ACKs for top commit:
achow101:
ACK 05117e6e17
ryanofsky:
Code review ACK 05117e6e17, just updated a commit message since last review
TheCharlatan:
ACK 05117e6e17
vasild:
ACK 05117e6e17
Tree-SHA512: 277c285a6e73dfff88fd379298190b264254996f98b93c91c062986ab35c2aa5e1fbfec4cd71d7b29dc2d68e33f252b5cfc501345f54939d6bd78599b71fec04
fa21f83d29 ci: Use G++ in valgrind tasks (MarcoFalke)
fabd05bf65 refactor: Fix net_processing iwyu includes (MarcoFalke)
fa1622db20 refactor: Make node_id a const& in RemoveBlockRequest (MarcoFalke)
Pull request description:
Currently, `valgrind` is not usable on a default build with GCC. Specifically, `p2p_compactblocks.py --valgrind` gives a false-positive in `RemoveBlockRequest` when comparing `node_id` with `from_peer`. According to the upstream bug report, this happens because both symbols are on the stack and the compiler can more aggressively optimize the compare (order). See https://bugs.kde.org/show_bug.cgi?id=472329#c7
It is possible to work around this bug by pulling at least one value from the stack. For example, by making `from_peer` a `const` reference. Alternatively, by replacing `auto [node_id, list_it]` with `const auto& [node_id, list_it]`, which is done here.
I think this workaround is acceptable, because it does not look like valgrind can trivially fix this. The alternative would be to add a (temporary?) suppression.
Fixes https://github.com/bitcoin/bitcoin/issues/27741
Also, fix iwyu includes, while touching this module.
Also, switch the CI valgrind scripts to use G++.
ACKs for top commit:
achow101:
ACK fa21f83d29
TheCharlatan:
ACK fa21f83d29
darosior:
utACK fa21f83d29
ryanofsky:
Code review ACK fa21f83d29. Code changes all look good but I'm a little confused about purpose of the third commit, so left a question about that
Tree-SHA512: 7b92cdafd525a5ac53ae2c1a7a92e599bc9b5fd5d315a694b493cd5079ac323d884393b57aa18581b7789247a588c9a27d47698de25b340bc76fc9f1dd1850b4
The obfuscation (XOR) operations are currently done byte-by-byte during serialization. Buffering the reads will enable batching the obfuscation operations later.
Different operating systems handle file caching differently, so reading larger batches (and processing them from memory) is measurably faster, likely because of fewer native fread calls and reduced lock contention.
Note that `ReadRawBlock` doesn't need buffering since it already reads the whole block directly.
Unlike `ReadBlockUndo`, the new `ReadBlock` implementation delegates to `ReadRawBlock`, which uses more memory than a buffered alternative but results in slightly simpler code and a small performance increase (~0.4%). This approach also clearly documents that `ReadRawBlock` is a logical subset of `ReadBlock` functionality.
The current implementation, which iterates over a fixed-size buffer, provides a more general alternative to Cory Fields' solution of reading the entire block size in advance.
Buffer sizes were selected based on benchmarking to ensure the buffered reader produces performance similar to reading the whole block into memory. Smaller buffers were slower, while larger ones showed diminishing returns.
------
> macOS Sequoia 15.3.1
> C++ compiler .......................... Clang 19.1.7
> cmake -B build -DBUILD_BENCH=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ && cmake --build build -j$(nproc) && build/bin/bench_bitcoin -filter='ReadBlockBench' -min-time=10000
Before:
| ns/op | op/s | err% | total | benchmark
|--------------------:|--------------------:|--------:|----------:|:----------
| 2,271,441.67 | 440.25 | 0.1% | 11.00 | `ReadBlockBench`
After:
| ns/op | op/s | err% | total | benchmark
|--------------------:|--------------------:|--------:|----------:|:----------
| 1,738,971.29 | 575.05 | 0.2% | 10.97 | `ReadBlockBench`
------
> Ubuntu 24.04.2 LTS
> C++ compiler .......................... GNU 13.3.0
> cmake -B build -DBUILD_BENCH=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++ && cmake --build build -j$(nproc) && build/bin/bench_bitcoin -filter='ReadBlockBench' -min-time=20000
Before:
| ns/op | op/s | err% | ins/op | cyc/op | IPC | bra/op | miss% | total | benchmark
|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
| 6,895,987.11 | 145.01 | 0.0% | 71,055,269.86 | 23,977,374.37 | 2.963 | 5,074,828.78 | 0.4% | 22.00 | `ReadBlockBench`
After:
| ns/op | op/s | err% | ins/op | cyc/op | IPC | bra/op | miss% | total | benchmark
|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
| 5,771,882.71 | 173.25 | 0.0% | 65,741,889.82 | 20,453,232.33 | 3.214 | 3,971,321.75 | 0.3% | 22.01 | `ReadBlockBench`
Co-authored-by: maflcko <6399679+maflcko@users.noreply.github.com>
Co-authored-by: Ryan Ofsky <ryan@ofsky.org>
Co-authored-by: Martin Leitner-Ankerl <martin.ankerl@gmail.com>
Co-authored-by: Cory Fields <cory-nospam-@coryfields.com>
Made every OpenBlockFile#fReadOnly value explicit.
Replaced hard-coded values in ReadRawBlock with STORAGE_HEADER_BYTES.
Changed `STORAGE_HEADER_BYTES` and `UNDO_DATA_DISK_OVERHEAD` to `uint32_t` to avoid casts.
Also added `LIFETIMEBOUND` to the `AutoFile` parameter of `BufferedFile`, which stores a reference to the underlying `AutoFile`, allowing Clang to emit warnings if the referenced `AutoFile` might be destroyed while `BufferedFile` still exists.
Without this attribute, code with lifetime violations wouldn't trigger compiler warnings.
Co-authored-by: maflcko <6399679+maflcko@users.noreply.github.com>
