9f8e7b0b3b node: cap -dbcache to 1GiB on 32-bit architectures (Antoine Poinsot)
2c43b6adeb init: cap -maxmempool to 500 MB on 32-bit systems (Antoine Poinsot)
Pull request description:
32-bit architecture is limited to 4GiB of RAM, so it doesn't make sense to set a too high value. A too high value could cause an OOM unbeknownst to the user a while after startup as mempool / dbcache fills.
ACKs for top commit:
achow101:
ACK 9f8e7b0b3b
instagibbs:
utACK 9f8e7b0b3b
dergoegge:
Code review ACK 9f8e7b0b3b
glozow:
utACK 9f8e7b0b3b
Tree-SHA512: cc7541b2c0040fc21a43916caec464dfb443af808f4e85deffa1187448ffff6edb0d69f9ebdb43915d145b8b4694d8465afe548f88da53ccebc9ce4b7c34b735
173394d951 depends: Build `qt` package for FreeBSD hosts (Hennadii Stepanov)
Pull request description:
This PR continues the work started in https://github.com/bitcoin/bitcoin/pull/23948.
Here is an excerpt from the log:
```
$ ./build/bin/bitcoin-qt -printtoconsole
2025-06-12T01:06:56Z Bitcoin Core version v29.99.0-15de25ba2a28 (release build)
2025-06-12T01:06:56Z Qt 6.7.3 (static), plugin=xcb
2025-06-12T01:06:56Z Static plugins:
2025-06-12T01:06:56Z QMinimalIntegrationPlugin, version 395008
2025-06-12T01:06:56Z QXcbIntegrationPlugin, version 395008
2025-06-12T01:06:56Z Style: fusion / QFusionStyle
2025-06-12T01:06:56Z System: FreeBSD 14.3-RELEASE, x86_64-little_endian-lp64
```
And here are the screenshots:
ACKs for top commit:
vasild:
ACK 173394d951
Tree-SHA512: 42a0bd11e4ef1a23efcfe6c4ab179dc667a076e65060891ce8358b3fe78de4e3ea33f975387d4236cc2ac620e2935b0a29c278065a47f038c66658106bf36755
8800b5acc1 cmake: Explicitly specify `Boost_ROOT` for Homebrew's package (Hennadii Stepanov)
Pull request description:
On macOS, this PR ensures that the Boost package is located at its real path rather than via the symlink in the default prefix.
A backport to 29.x is required for https://github.com/bitcoin/bitcoin/pull/32804, as this change prevents contamination of include directories by broad locations such as `/usr/local/include` or `/opt/homebrew/include`, which take precedence over Qt’s `-iframework` flags.
Below is the relevant change in the configuration logs on my macOS 15.5 `x64`:
- master branch @ ead4468748:
```
% cmake -B build
<snip>
-- Found Boost: /usr/local/include (found suitable version "1.88.0", minimum required is "1.73.0")
<snip>
```
- this PR:
```
% cmake -B build
<snip>
-- Found Boost: /usr/local/opt/boost/include (found suitable version "1.88.0", minimum required is "1.73.0")
<snip>
```
This PR is forward-compatible with the changes proposed in https://github.com/bitcoin/bitcoin/pull/32667.
ACKs for top commit:
fanquake:
ACK 8800b5acc1 Checked that this plus #32805fixes#31009
Tree-SHA512: 114bd945ec0c06a8d15b565e5b9aafc3bcfdf2a4ba4400e072b8e31053dff0f9ef423b941ee1dff2113f83e08f7fada728383ae88b3ec380b5c3e40553205f7d
6c2538d5bf depends: Bump boost to 1.88.0 and use new CMake buildsystem (Cory Fields)
Pull request description:
Originally #30434.
This has a few advantages over the old method of simply copying headers:
- Installs proper CMake files which can be picked up by our buildsystem
- Only installs necessary headers, not all of Boost
Pulls in upstreamed https://github.com/boostorg/test/pull/445.
ACKs for top commit:
willcl-ark:
tACK 6c2538d5bf
hebasto:
re-ACK 6c2538d5bf, only rebased since my previous [review](https://github.com/bitcoin/bitcoin/pull/32665#pullrequestreview-2891203225).
Tree-SHA512: fc3fce77b21c8ea500370841f44f1cc87e0bb09cdde55f75d2f90853cb06a6f8c73ac6ca9ca3e91a879e9f95dd59aa40254c1b04e7a168c52fa1b31cc5b7f537
ead4468748 cmake: Use `HINTS` instead of `PATHS` in `find_*` commands (Hennadii Stepanov)
Pull request description:
According to the CMake documentation, `HINTS` "should be paths computed by system introspection, such as a hint provided by the location of another item already found", which is precisely the case in the `FindQRencode` module.
Entries in `HINTS` are searched before those in `PATHS`. On macOS, Homebrew’s `libqrencode` will therefore be located at its real path rather than via the symlink in the default prefix.
A backport to 29.x is required for https://github.com/bitcoin/bitcoin/pull/32804, as this change prevents contamination of include directories by broad locations such as `/usr/local/include` or `/opt/homebrew/include`, which take precedence over Qt’s `-iframework` flags.
Below is the relevant change in the configuration logs on my macOS 15.5 `x64`:
- master branch @ ead4468748:
```
% cmake -B build -DBUILD_GUI=ON
<snip>
-- Found QRencode: /usr/local/lib/libqrencode.dylib (found version "4.1.1")
<snip>
```
- this PR:
```
% cmake -B build -DBUILD_GUI=ON
<snip>
-- Found QRencode: /usr/local/Cellar/qrencode/4.1.1/lib/libqrencode.dylib (found version "4.1.1")
<snip>
```
ACKs for top commit:
fanquake:
ACK ead4468748
Tree-SHA512: 1f0b04e3efeb7fe3efbb969be911abbcf56030d715acd87c0fbaf24422cdf1122f169e32242571256916c96a159212842e1e73092145c63ecc495ce429c6e587
- Increase block weight by 4000 for all nodes with custom -blockmaxweight.
Prior to this commit, we generated blocks with 4000 weight units less worth of transactions.
See https://github.com/bitcoin/bitcoin/issues/32461#issuecomment-2925282272 for details.
This commit fixes it by increasing the block weight by 4000.
- Update `check_smart_estimates` to calculate the fee rate ceiling
by taking the maximum of fees seen, minrelaytxfee, and mempoolminfee.
- Improve the subtest name and comments.
dd8447f70f test: fix catchup loop in outbound eviction functional test (Sebastian Falbesoner)
Pull request description:
In the course of working on an equivalent of #32421 for the `CBlockHeader` class, I noticed that the [catchup loop in the outbound eviction functional test](19765dca19/test/functional/p2p_outbound_eviction.py (L86-L103)) currently has a small flaw: the contained waiting for a `getheaders` message
19765dca19/test/functional/p2p_outbound_eviction.py (L98-L99)
only waits for _any_ such message instead of one with the intended block hash after the first iteration. The reason is that the `prev_prev_hash` variable is set incorrectly, since the `tip_header` instance is not updated and its field `.hash` is None [1]. Fix that by updating `tip_header` after generating a new block and also use the correct field on it -- we want the tip header's previous hash (`.hashPrevBlock`), which will be the previous-previous hash in the next iteration as intended.
Can be demonstrated by adding a debug output for `prev_prev_hash`, e.g.
```diff
diff --git a/test/functional/p2p_outbound_eviction.py b/test/functional/p2p_outbound_eviction.py
index 30ac85e32f..9886a49512 100755
--- a/test/functional/p2p_outbound_eviction.py
+++ b/test/functional/p2p_outbound_eviction.py
@@ -85,6 +85,7 @@ class P2POutEvict(BitcoinTestFramework):
self.log.info("Keep catching up with the old tip and check that we are not evicted")
for i in range(10):
+ print(f"i={i}, prev_prev_hash={prev_prev_hash}")
# Generate an additional block so the peers is 2 blocks behind
prev_header = from_hex(CBlockHeader(), node.getblockheader(best_block_hash, False))
best_block_hash = self.generateblock(node, output="raw(42)", transactions=[])["hash"]
```
master branch
```
...
i=0, prev_prev_hash=21722572577213525620063947414919931742473663114977483853465070858884938201585
i=1, prev_prev_hash=None
i=2, prev_prev_hash=None
i=3, prev_prev_hash=None
i=4, prev_prev_hash=None
i=5, prev_prev_hash=None
i=6, prev_prev_hash=None
i=7, prev_prev_hash=None
i=8, prev_prev_hash=None
i=9, prev_prev_hash=None
...
```
PR branch
```
...
i=0, prev_prev_hash=21722572577213525620063947414919931742473663114977483853465070858884938201585
i=1, prev_prev_hash=23204083306104595181276643925327085197417756603258684897360269464191973063397
i=2, prev_prev_hash=18117007775254206852722585270408843074799046031613422902091537272077477361634
i=3, prev_prev_hash=30556804635951812756130312631227721973553160707632138130845362630877961299882
i=4, prev_prev_hash=16476515948153779819467376247405243058769281687868039119037064816106574626111
i=5, prev_prev_hash=14965506521435221774966695805624206855826023174786191695076697927307467053159
i=6, prev_prev_hash=14510815979277079515923749862202324542606166669768865640616202929053689167149
i=7, prev_prev_hash=15360268707191667685151951417759114642582372006627142890517655217275478262166
i=8, prev_prev_hash=55984929479619644661389829786223559362979512070332438490054115824374865094074
i=9, prev_prev_hash=6591573629906616262191232272909118561529534571119028248829355592878183757083
...
```
[1] that's in my opinion another example how caching hashes is confusing and easy to be misused; it's better to remove it and just compute the hash on-the-fly, so returning None is not even possible anymore
ACKs for top commit:
maflcko:
lgtm ACK dd8447f70f
mzumsande:
Code Review ACK dd8447f70f
pablomartin4btc:
cr-ACK dd8447f70f
Tree-SHA512: bd8e786b52e3e96661453006140d6b8fad5a35f1c8d38243c61df52b19c97cd3800404745a2f9603bcdf0006e9780b4f15f8f7e4fa34ff07d52dba04d87b68d0
When a legacy wallet has been migrated to contain descriptors, but
before the transactions have been updated to match, we need to recompute
the wallet TXOs so that the transaction update will work correctly.
Instead of searching mapWallet for the preselected inputs, search
m_txos.
wallet_fundrawtransaction.py spends external inputs and needs the change
output to also belong to the test wallet for the oversized tx test.
Instead of iterating every transaction and every output stored in wallet
when trying to figure out what outputs can be spent, iterate the TXO set
which should be a lot smaller.
Since we track the outputs owned by the wallet with m_txos, we can now
calculate the balance of the wallet by iterating m_txos and summing up
the amounts of the unspent txos.
As ISMINE_USED is not an actual isminetype that we attach to outputs and
was just passed into `CachedTxGetAvailableCredit` for convenience, we
pull the same determining logic from that function into `GetBalances` in
order to preserve existing behavior.
After adding a wallet descriptor (typically by import), mark all balance
caches dirty. This allows transactions that the wallet already knows
about that have outputs that are now ISMINE_SPENDABLE after the import
to actually be shown in balance calculations. Legacy wallet imports
would do this, but importdescriptors did not.
One of the main issues with AvailableCoins is its performance when txs
have unrelated outputs, so update the benchmark to check the performance
of that.
c3fe85e2d6 wallet, rpc, test: Remove deprecated getunconfirmedbalance (Ava Chow)
0ec255139b wallet, rpc: Remove deprecated balances from getwalletinfo (Ava Chow)
Pull request description:
`getwalletinfo` result fields `balance`, `immature_balance`, and `unconfirmed_balance`, and the `getunconfirmedbalance` RPC have all been deprecated since 0.19.0. It's been long enough that they should either be removed or undeprecated. The functionality provided by these RPCs is provided by `getbalances`.
ACKs for top commit:
davidgumberg:
ACK c3fe85e2d6
rkrux:
ACK c3fe85e2d6
BrandonOdiwuor:
ACK c3fe85e2d6 removing the deprecated `balance, unconfirmed_balance, immature_balance` fields from `getwalletinfo` and `getunconfirmedbalance` RPCs, as this infomation can be found on the `getbalances` RPC
w0xlt:
reACK c3fe85e2d6
Tree-SHA512: c7c4acfd9cabc7517ba813b95281a6c6a717a417312afd9346298669b4f7bd37724ad977148ce42db7fd47fc3d1f5a8482d8ff2e7b9cb74756b171a5b8b91ef2
47237cd193 wallet, rpc: Output wallet flags in getwalletinfo (Ava Chow)
bc2a26b296 wallet: Add GetWalletFlags (Ava Chow)
69f588a99a wallet: Set upgraded descriptor cache flag for newly created wallets (Ava Chow)
Pull request description:
Newly created wallets will always have an upgraded descriptor cache, so set those.
Also, to verify this behavior, add a new `flags` field to `getwalletinfo` and check that in the functional tests.
Split from #32489
ACKs for top commit:
Sjors:
ACK 47237cd193
w0xlt:
ACK 47237cd193
rkrux:
ACK 47237cd193
Tree-SHA512: 97c7f85b858efe5ced9b8aafb6cd7c1a547de6f8013b82bfc75bc567cf73c9db5e168e3980355756541305520022fd776b8d4d240d3fb34ed86c27d2acaf4863
The wallet backups performed before migration use the time as part of
their filename. As the time is mocked, increment it between migration
attempts to prevent file name conflicts which is a problem on Windows.
This test required v0.14.3 which doesn't support special characters like
emojis in the datapath on windows. Functional test runner includes a few
emojis in it's default data directory.
9eb2c82e7c walletdb: Remove unused upgraded_txs (Ava Chow)
c668033709 wallet: Remove unused fTimeReceivedIsTxTime (Ava Chow)
Pull request description:
`CWalletTx::fTimeReceivedIsTxTime` is no longer used and can be removed. This additionally allows the removal of the `upgraded_txs` loop in `LoadWallet`.
ACKs for top commit:
maflcko:
lgtm ACK 9eb2c82e7c
Eunovo:
ACK 9eb2c82e7c
davidgumberg:
ACK 9eb2c82e7c
PeterWrighten:
ACK 9eb2c82e7c
rkrux:
ACK 9eb2c82e7c
w0xlt:
ACK 9eb2c82e7c
Tree-SHA512: 05cf3a50f0d8ab6ef423ad1113c5ce6f45bfdc90e2c0dcf61c2dceced2465502e574b4b5b0091fcbb4bdd2182f8d69224f1e5516c7c505de07102b84a5f40e9c
Windows zip files are added in the next commit which are not tarballs so
renaming tarball to the more generic term archive which can cover both.
-BEGIN VERIFY SCRIPT-
sed -i 's/tarball/archive/g' test/get_previous_releases.py
-END VERIFY SCRIPT-
Using the get_previous_releases.py script to build from source only works for
releases prior to v29 due to removal of Autotools (in favor of CMake). It also
does not support building on Windows, and we are adding support for downloading
Windows release binaries in later commits of this PR.
As there were no complaints during review, it is assumed nobody uses this
functionality.
Current recommendation is to add a new remote fetching all PRs, but this
is resource-intensive.
Document a better way to fetch a single PR, and to update a PR which has
been force-pushed.
This has a few advantages over the old method of simply copying headers:
- Installs proper cmake files which can be picked up by our buildsystem
- Only installs necessary headers, not all of boost
Co-authored-by: Hennadii Stepanov <32963518+hebasto@users.noreply.github.com>
e27a94596f build: add root dir to CMAKE_PREFIX_PATH (will)
Pull request description:
Fixes: #32428
Nix patches `cmake` to remove the root directory `/` from `CMAKE_PREFIX_PATH`:
428b49b28e/pkgs/by-name/cm/cmake/001-search-path.diff (L10)
Without this, and when using the toolchain for depends builds, cmake's `find_path()` and `find_package()` do not know where to find dependencies, causing issues like as seen in #32428
Adding this path back is harmless on other systems, and fixes the toolchain for Nix users.
As described in https://github.com/bitcoin/bitcoin/issues/32428#issuecomment-2991258328 I think this can be taken as a temporary fix whilst a longer-term solution is worked on.
ACKs for top commit:
josibake:
reACK e27a94596f
hebasto:
ACK e27a94596f, I have reviewed the code and it looks OK.
janb84:
reACK e27a94596f
Tree-SHA512: f299f2bab2620179518da866cbb7992d41d142ad42e79c14496e72f725a1dc60698b3e4b1daf45d28f71f32a23f0c8d7b4f6c6cf33aeedf322b7ef565b70b4af
According to the CMake documentation, `HINTS` "should be paths computed
by system introspection, such as a hint provided by the location of
another item already found", which is precisely the case in the
`FindQRencode` module.
Entries in `HINTS` are searched before those in `PATHS`. On macOS,
Homebrew’s `libqrencode` will therefore be located at its real path
rather than via the symlink in the default prefix.
fa68dcb207 ci: Add missing errexit to lint CI install (MarcoFalke)
fa535a6de7 ci: Allow running CI in worktrees (MarcoFalke)
faf6a04597 ci: Clean UID/GID mismatch (MarcoFalke)
Pull request description:
Fixes#30028 (modulo lint and tidy CI).
The error on current master in a worktree is:
```
$ git worktree add ./main origin/master && cd ./main
$ MAKEJOBS="-j$(nproc)" FILE_ENV="./ci/test/00_setup_env_mac_cross.sh" ./ci/test_run_all.sh
...
+ git config --global ci.base-install-done true
fatal: not a git repository: /root/b-c-ci/.git/worktrees/main
```
So just use a plain file, instead of git. Also, enable pipefail while touching this bash script.
ACKs for top commit:
willcl-ark:
tACK fa68dcb207
Tree-SHA512: 0ce360a80883b4aa655fe8a99c38eb54a465b17c7cdb0a69a2d886ff78da32d6af996412ffc5b0db0322acafa9650619838573484de8243dc41594a04a6e17ec
Nix patches cmake to remove the root directory `/` from
`CMAKE_SYSTEM_PREFIX_PATH`:
428b49b28e/pkgs/by-name/cm/cmake/001-search-path.diff (L10)
Without this, and when using the toolchain for depends builds, cmake's
`find_path()` and `find_package()` do not know where to find
dependencies, causing issues like:
https://github.com/bitcoin/bitcoin/issues/32428
Adding this path back via CMAKE_PREFIX_PATH is harmless on other
systems, and fixes the toolchain for Nix users.
We append the `/` dir a maximum of once, as the toolchain may be called
repeatedly during builds.
Co-authored-by: Russell Yanofsky <russ@yanofsky.org>
Co-authored-by: josibake <josibake@protonmail.com>
Reason for each test:
rpc_whitelist.py: Relies on direct RPC calls
wallet_encryption.py: Null characters cannot be passed to suprocess.Popen
wallet_fundrawtransaction.py: multiple checks for wrong types, which have different error messages with cli
wallet_send.py: multiple checks for wrong types
