d71e29e3e828bcb7b702fad728546351b8db5c01 qa: Correct epoll_ctl data race suppression (Hennadii Stepanov)
Pull request description:
Fixup of #20218. Comments must start from the beginning of the line.
ACKs for top commit:
MarcoFalke:
review ACK d71e29e3e828bcb7b702fad728546351b8db5c01
Tree-SHA512: 4d8663ab505c347bcb62c2f118656e3343d5179825be0d1b86761ffdfdae1e7462002bf226a54dfc94be5885ce7f2633abaf70421ea35bf06eddad8e99fb9683
02ac445b2fec60e028d2cc93bbf74a35e3d0f48e bump libevent to 2.1.11 in depends (stefanwouldgo)
Pull request description:
this doesn't need patches on Android anymore like 2.1.8 did.
ACKs for top commit:
laanwj:
ACK 02ac445b2fec60e028d2cc93bbf74a35e3d0f48e
Tree-SHA512: 1fbfe342ee15fa4c5cb417979bd6c443f7c7aa40a489accf8ccd7c919e5b08e859b3da6edeee3de484f6f156b35dd4e97c7e2c7971b59fc31029865585ccb296
880d4aaf81f3d5d7fbb915905c2e61b816a6a747 build: use BOOST_NO_CXX98_FUNCTION_BASE to suppress warnings (fanquake)
1bdbbbdc46c4e50bf07bc362e7e391ea1a53ea2f build: suppress array-bounds errors in libxkbcommon (fanquake)
Pull request description:
2 changes to better support building with GCC 12, which out of the box, is currently broken if you want to build using depends.
Prevent `-Warray-bounds` errors when building libxkbcommon. i.e:
```bash
src/xkbcomp/ast-build.c:82:27: error: array subscript 'ExprDef[0]' is partly outside array bounds of 'unsigned char[32]' [-Werror=array-bounds]
82 | expr->expr.value_type = type;
| ~~~~~~~~~~~~~~~~~~~~~~^~~~~~
src/xkbcomp/ast-build.c:75:21: note: object of size 32 allocated by 'malloc'
75 | ExprDef *expr = malloc(size);
| ^~~~~~~~~~~~
```
It might be the case that these would be fixed by updating the
package, but that would also require installing new build tools (meson),
as well as potentially more dependencies (wayland), and it'd need
testing with Qt. For now, just turn the errors into wanrings.
Define `BOOST_NO_CXX98_FUNCTION_BASE` to prevent GCC warning about the use of `std::unary_function`. i.e:
```bash
/bitcoin/depends/aarch64-unknown-linux-gnu/include/boost/container_hash/hash.hpp:131:33:
warning: 'template<class _Arg, class _Result> struct std::unary_function' is deprecated [-Wdeprecated-declarations]
131 | struct hash_base : std::unary_function<T, std::size_t> {};
| ^~~~~~~~~~~~~~
In file included from /usr/include/c++/12/bits/unique_ptr.h:37,
from /usr/include/c++/12/memory:76,
from ./init.h:10,
from init.cpp:10:
/usr/include/c++/12/bits/stl_function.h:117:12: note: declared here
117 | struct unary_function
```
Boost `container_hash` (included via functional -> multi_index) uses
[`std::unary_function`, which was deprecated in C++11](https://en.cppreference.com/w/cpp/utility/functional/unary_function), and "removed" in
C++17. It's use causes warnings with newer compilers, i.e GCC 12.1.
Use the MACRO outlined in https://github.com/boostorg/container_hash/issues/22, and added to Boost Config for GCC 12 in https://github.com/boostorg/config/pull/430, to prevent it's use.
[BOOST_NO_CXX98_FUNCTION_BASE](https://www.boost.org/doc/libs/master/libs/config/doc/html/boost_config/boost_macro_reference.html):
> The standard library no longer supports std::unary_function and std::binary_function.
> They were deprecated in C++11 and is removed from C++14.
Guix Build (x86_64):
```bash
```
Guix Build (arm64):
```bash
```
ACKs for top commit:
laanwj:
Code review ACK 880d4aaf81f3d5d7fbb915905c2e61b816a6a747
Tree-SHA512: 10c4679c3eb788e9279acc4960731c55ae1568bd3df525d3c46f97d8b0319e7d8450b1638b6777d98111b5991dba5c787e95d80b1ac932e0b4779d4b8e74875e
Unfortunately, we need -Wno-unused-command-line-argument as configure will fail to run
with -Werror due to Dash's current build system utilizing -static-libstdc++ indiscriminately,
resulting in Clang warnings of "unused arguments".
* llmq only logging
* llmq only logging
* style: reference on the left
* refactoring
* fix
* style: fix colon location in for loop
Co-authored-by: PastaPastaPasta <6443210+PastaPastaPasta@users.noreply.github.com>
facb71576cd4d2e90fd03e09d29b42fa3d730e8c net: Remove forcerelay of rejected txs (MarcoFalke)
Pull request description:
This removes the code that supposedly handled the forced relay of txs from a permissioned peer that were rejected from our mempool. The removal should be fine, because it is dead code for the following reasons:
* While `RelayTransaction` enqueues the inv for all peers, the inv is never processed because it can not be found in the mempool. See 4a07233076/src/net_processing.cpp (L3862-L3866)
* Even if the peers we intended to send the inv to can somehow reply with a getdata to the never-received inv, they won't receive the tx as a reply because it was never added to the "relay memory" (`mapRelay`)
The dead code is (obviously) untested: https://marcofalke.github.io/btc_cov/total.coverage/src/net_processing.cpp.gcov.html#2574
This feature was (intentionally or accidentally) removed in 4d8993b346, which was released in Bitcoin Core 0.13.0. So all currently supported versions of Bitcoin Core ship without this feature. I am not aware of any complaints about this feature or actual documented use-cases. So instead of reviving an unneeded feature, just remove the dead code.
ACKs for top commit:
hebasto:
ACK facb71576cd4d2e90fd03e09d29b42fa3d730e8c, locally running the unit and functional tests.
Tree-SHA512: bfceae6f2983c1510fa0649a9a63c343cbbc1c4ab3a3698039cccf454c81e58c8f5114b147ed42a1bc867da74c43a5b53764ab14f942e191b6f59079044108b5
aaaae4d0ebd5ef34d81997a73ab9839ba7b4b9e4 test: Add p2p test for forcerelay permission (MarcoFalke)
fa6b57bcaaf4dc65d78316353033b03d171a3beb test: Fix whitespace in p2p_permissions.py (MarcoFalke)
faf40810d7b7f42f3588bfa8a663095aa24001b1 test: Make msg_tx a witness tx (MarcoFalke)
Pull request description:
The commit `test: Make msg_tx a witness tx` is needed so that the python mininode does not strip the witness from transactions before sending them over p2p. The commit should also be done to keep symmetry with msg_block. See:
* tests: Make msg_block a witness block #15982
ACKs for top commit:
laanwj:
ACK aaaae4d0ebd5ef34d81997a73ab9839ba7b4b9e4
Tree-SHA512: b4b546c88f7f0576cb512f0872bc6bef9d4df65783803f226986e56175937f418aa1ed906417ac909f27f1fd521d64629621fda83250fa925c46ef9513db0e4c
5ffaf883b93fb8d3d841c36e808b90d61d39d492 test: eliminiated magic numbers in feature_csv_activation.py (Sebastian Falbesoner)
09f706ab8e47ddfdfa41418f2e7cb6d87661147a test: check for OP_CSV empty stack fail reject reason in feature_csv_activation.py (Sebastian Falbesoner)
cbd345a75c2be26e17fce4c65c0c1ca19a3eb9e0 test: test OP_CSV empty stack fail in feature_csv_activation.py (Sebastian Falbesoner)
Pull request description:
Adds an empty stack failure check for OP_CSV (BIP112) to the functional test `feature_csv_activation.py` by prepending a valid scriptSig with `OP_CHECKSEQUENCEVERIFY`.
If BIP112 is inactive, the operator just behaves as a NOP (for both tx versions 1 and 2) and the transaction remains valid -- if it is active, the tx is invalid due to an empty stack (for both tx versions 1 and 2, as well).
Top commit has no ACKs.
Tree-SHA512: 81102aaead5be11e02b894867fa9a9cc17358ec0eb2f21ce2d3db845b87691d305e6ed7c525f9c7e5bcb3c5c609eb28deca0fbaa3d5e9ff928cecd3b91ff129a
fa45d606461dbf5bf1017d6ab15e89c1bcf821a6 test: Reduce unneeded whitelist permissions in tests (MarcoFalke)
Pull request description:
It makes the tests confusing and fragile when overwriting default command line values that are not needed to be overwritten.
ACKs for top commit:
fanquake:
ACK fa45d606461dbf5bf1017d6ab15e89c1bcf821a6
laanwj:
ACK fa45d606461dbf5bf1017d6ab15e89c1bcf821a6
Tree-SHA512: 8ae5ad8c6be156b1a983adccbca8d868ef841e00605ea88e24227f1b7493987c50b3e62e68dd7dc785ad73d6e14279eb13d7a151cb0a976426fe2fd63ce5cbcd
b902bd66b0f35c5016dc5d7aaf501940935edd62 test: check custom descendant limit in mempool_packages.py (Sebastian Falbesoner)
Pull request description:
This is a follow-up PR to #17435, testing the custom descendant limit, passed by the argument `-limitdescendantcount`. ~~It was more tricky than expected, mainly because we don't know for sure at which point node1 has got all the transactions broadcasted from node0 (for the ancestor test this wasn't a problem since the txs were immediately available through `invalidateblock`) -- a simple `sync_mempools()` doesn't work here since the mempool contents are not equal due to different ancestor/descendant limits. Hence I came up with a "hacky manual sync":~~
1. ~~wait until the mempool has the _expected_ tx count (see conditions below)~~
2. ~~after that, wait some time and get sure that the mempool contents haven't changed in-between~~
~~Like for~~ Similar to the ancestor test, we overall check for ~~three~~ four conditions:
- the # of txs in the node1 mempool is equal to the descendant limit (plus 1 for the parent tx, plus the # txs from the previous ancestor test which are still in) ~~(done by the hacky sync above)~~
- all txs in node1 mempool are a subset of txs in node0 mempool
- part of the constructed descendant-chain (the first ones up to the limit) are contained in node1 mempool
- the remaining part of the constructed descendant-chain (all after the first ones up to the limit) is *not* contained in node1 mempool
ACKs for top commit:
JeremyRubin:
Excellent. utACK b902bd6
Tree-SHA512: 7de96dd248f16ab740e178ac5b64b57ead18cdcf74adfe989709d215e4a67b6b6d20de22c48e885d5f2edc55caaddd44a4261e996c5c87687ceb6a47f1d1fdaf
0653939ac130eddffe40c53ac418bea305d3bf82 Add static_asserts to ser_X_to_Y() methods (Samer Afach)
be94096dfb0c4862e2314cbae4120d7360b08ef2 Fix a violation of C++ standard rules that unions cannot be switched. (Samer Afach)
Pull request description:
Type punning in C++ is not like C. As per the C++ standard, one cannot use unions to convert the bit type. A discussion about this can be found [here](https://stackoverflow.com/questions/25664848/unions-and-type-punning). In C++, a union is supposed to only hold one type at a time. It's intended to be used only as `std::variant`. Switching types is undefined behavior.
In fact, C++20 has a special casting function, called [`bit_cast`](https://en.cppreference.com/w/cpp/numeric/bit_cast) that solved this problem.
Why has it been working so far? Because some compilers tolerate using unions and switching types, like gcc. More information [here](https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#Type-punning).
One important thing to mention is that performance is generally not affected by that memcpy. Compilers are smart enough to convert that to a memory cast when possible. But we have to do it the right way, otherwise, it's jut undefined behavior that depends on the compiler.
ACKs for top commit:
practicalswift:
ACK 0653939ac130eddffe40c53ac418bea305d3bf82
elichai:
ACK 0653939ac130eddffe40c53ac418bea305d3bf82
laanwj:
Code review ACK 0653939ac130eddffe40c53ac418bea305d3bf82
kristapsk:
ACK 0653939ac130eddffe40c53ac418bea305d3bf82
Tree-SHA512: f6e89de39fc964750429139bab6b5a1346f7060334b7afa020e315bdad8f8c195bce2b8a9e343f06e7fff175e2dfb1cdabfcb6fe405bea0febe4962f0cc62557
5bad7921d0b33b62c0a59a478c2e8c869fc5e3b5 [test] PSBT RPC: check that bip32_derivs are present by default (Sjors Provoost)
29a21c90610aed88b796a7a5900e42e9048b990e [rpc] set default bip32derivs to true for psbt methods (Sjors Provoost)
Pull request description:
In https://github.com/bitcoin/bitcoin/pull/13557#pullrequestreview-135905054 I recommended not including bip32 deriviation by default in PSBTs:
> _Bit of a privacy issue_: let's say person A and B are about to spend from a multisig address, sending everything to person A. Person A gives their address to person B, their wallet wallet creates a PSBT, but doesn't sign it. Wallet A then calls `walletprocesspsbt` which signs it and _spontaneously adds the master_fingerprint and bip32 path_. Same issue with `walletcreatefundedpsbt`.
>
> Adding `bip32_derivs` should probably be opt-in.
In practice I find this default quite annoying because I forget it and end up with a confused hardware wallet.
More importantly, in the multisig example I provided, it's actually essential for the other side to know the derivation details (in addition to an xpub). This allows them to check that change is going to an address you can still co-sign for (because the spending policy is unchanged except for an index).
ACKs for top commit:
instagibbs:
utACK 5bad7921d0
jonatack:
ACK 5bad7921d0 code review, built, ran tests, inspected/messed around with/pprinted values from the new tests. Thanks for adding the tests.
meshcollider:
utACK 5bad7921d0b33b62c0a59a478c2e8c869fc5e3b5
Tree-SHA512: 22ad71dda96856060a96758c4ae7aafa22d5e9efba30e0c8287c711e7579849bd72593cbc0f41a2e9e8821315d78bda04e848dbb006283b841b2795e2faebcfd
12a2f377185a413b740460db36812de22ee2e041 util: Avoid potential uninitialized read in FormatISO8601DateTime(int64_t nTime) by checking gmtime_s/gmtime_r return value (practicalswift)
Pull request description:
Avoid potential uninitialized read in `FormatISO8601DateTime(int64_t)` by checking `gmtime_s`/`gmtime_r` return value.
Before this patch `FormatISO8601DateTime(67768036191676800)` resulted in:
```
==5930== Conditional jump or move depends on uninitialised value(s)
==5930== at 0x4F44C0A: std::ostreambuf_iterator<char, std::char_traits<char> > std::num_put<char, std::ostreambuf_iterator<char, std::char_traits<char> > >::_M_insert_int<long>(std::ostreambuf_iterator<char, std::char_traits<char> >, std::ios_base&, char, long) const (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.25)
==5930== by 0x4F511A4: std::ostream& std::ostream::_M_insert<long>(long) (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.25)
==5930== by 0x4037C3: void tinyformat::formatValue<int>(std::ostream&, char const*, char const*, int, int const&) (tinyformat.h:358)
==5930== by 0x403725: void tinyformat::detail::FormatArg::formatImpl<int>(std::ostream&, char const*, char const*, int, void const*) (tinyformat.h:543)
==5930== by 0x402E02: tinyformat::detail::FormatArg::format(std::ostream&, char const*, char const*, int) const (tinyformat.h:528)
==5930== by 0x401B16: tinyformat::detail::formatImpl(std::ostream&, char const*, tinyformat::detail::FormatArg const*, int) (tinyformat.h:907)
==5930== by 0x4017AE: tinyformat::vformat(std::ostream&, char const*, tinyformat::FormatList const&) (tinyformat.h:1054)
==5930== by 0x401765: void tinyformat::format<int, int, int, int, int, int>(std::ostream&, char const*, int const&, int const&, int const&, int const&, int const&, int const&) (tinyformat.h:1064)
==5930== by 0x401656: std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > tinyformat::format<int, int, int, int, int, int>(char const*, int const&, int const&, int const&, int const&, int const&, int const&) (tinyformat.h:1073)
==5930== by 0x4014CC: FormatISO8601DateTime[abi:cxx11](long) (…)
```
The same goes for other very large positive and negative arguments.
Fix by simply checking the `gmtime_s`/`gmtime_r` return value :)
ACKs for top commit:
MarcoFalke:
ACK 12a2f377185a413b740460db36812de22ee2e041
theStack:
re-ACK 12a2f37718
elichai:
re ACK 12a2f377185a413b740460db36812de22ee2e041
Tree-SHA512: 066142670d9bf0944d41fa3f3c702b1a460b5471b93e76a619b1e818ff9bb9c09fe14c4c37e9536a04c99533f7f21d1b08ac141e1b829ff87ee54c80d0e61d48
