test: tighten `V2TransportTester::SendMessage` bounds for GCC 15.2

l0rinc commented at 6:49 pm on November 27, 2025: contributor

Summary

#33840 adjusted V2TransportTester::SendMessage to avoid a -Warray-bounds false positive with GCC 13, but GCC 15.2 still warns.

Fix

Implemented the suggestion in #33840#pullrequestreview-3447477940:

take mtype as std::string_view;
assert mtype.size() <= CMessageHeader::MESSAGE_TYPE_SIZE;
build [0x00][12-byte type][payload] via std::ranges::copy.

Reproducer

Before this change, using GCC 15.2 on a Mac

rm -rfd build && cmake -B build -DENABLE_IPC=OFF -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++ && cmake –build build -j$(nproc)

produces:

 0[100%] Building CXX object src/test/CMakeFiles/test_bitcoin.dir/__/wallet/test/walletload_tests.cpp.o
 1In file included from /opt/homebrew/Cellar/gcc/15.2.0/include/c++/15/array:45,
 2                 from /opt/homebrew/Cellar/gcc/15.2.0/include/c++/15/span:46,
 3                 from /Users/lorinc/bitcoin/src/span.h:10,
 4                 from /Users/lorinc/bitcoin/src/uint256.h:10,
 5                 from /Users/lorinc/bitcoin/src/consensus/params.h:10,
 6                 from /Users/lorinc/bitcoin/src/kernel/chainparams.h:9,
 7                 from /Users/lorinc/bitcoin/src/chainparams.h:9,
 8                 from /Users/lorinc/bitcoin/src/test/net_tests.cpp:5:
 9In function 'constexpr _OutIter std::__copy_move_a2(_InIter, _Sent, _OutIter) [with bool _IsMove = false; _InIter = char*; _Sent = char*; _OutIter = unsigned char*]',
10    inlined from 'constexpr _OI std::__copy_move_a1(_II, _II, _OI) [with bool _IsMove = false; _II = char*; _OI = unsigned char*]' at /opt/homebrew/Cellar/gcc/15.2.0/include/c++/15/bits/stl_algobase.h:492:42,
11    inlined from 'constexpr _OI std::__copy_move_a(_II, _II, _OI) [with bool _IsMove = false; _II = __gnu_cxx::__normal_iterator<char*, __cxx11::basic_string<char> >; _OI = __gnu_cxx::__normal_iterator<unsigned char*, vector<unsigned char> >]' at /opt/homebrew/Cellar/gcc/15.2.0/include/c++/15/bits/stl_algobase.h:500:31,
12    inlined from 'constexpr _OI std::copy(_II, _II, _OI) [with _II = __gnu_cxx::__normal_iterator<char*, __cxx11::basic_string<char> >; _OI = __gnu_cxx::__normal_iterator<unsigned char*, vector<unsigned char> >]' at /opt/homebrew/Cellar/gcc/15.2.0/include/c++/15/bits/stl_algobase.h:642:7,
13    inlined from 'void net_tests::{anonymous}::V2TransportTester::SendMessage(std::string, std::span<const unsigned char>)' at /Users/lorinc/bitcoin/src/test/net_tests.cpp:1304:18:
14/opt/homebrew/Cellar/gcc/15.2.0/include/c++/15/bits/stl_algobase.h:426:32: warning: 'void* __builtin_memmove(void*, const void*, long unsigned int)' offset [0, 1] is out of the bounds [0, 0] [-Warray-bounds=]
15  426 |               __builtin_memmove(_GLIBCXX_TO_ADDR(__result),
16      |               ~~~~~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~
17  427 |                                 _GLIBCXX_TO_ADDR(__first),
18      |                                 ~~~~~~~~~~~~~~~~~~~~~~~~~~
19  428 |                                 __n * sizeof(*__first));
20      |                                 ~~~~~~~~~~~~~~~~~~~~~~~
21[100%] Linking CXX executable ../../bin/test_bitcoin

test: tighten `V2TransportTester::SendMessage` bounds for GCC 15.2

#33840 adjusted V2TransportTester::SendMessage to avoid a -Warray-bounds false positive with GCC 13, but GCC 15.2 still warns because the helper does not state that the message type must fit into MESSAGE_TYPE_SIZE.

Make that contract explicit and simplify the copy:
- take mtype as std::string_view;
- assert mtype.size() <= CMessageHeader::MESSAGE_TYPE_SIZE;
- build `[0x00][12-byte type][payload]` via std::ranges::copy.

Behavior is unchanged; the remaining GCC warning goes away.

b3e2ae7aae

DrahtBot added the label Tests on Nov 27, 2025

DrahtBot commented at 6:49 pm on November 27, 2025: contributor

The following sections might be updated with supplementary metadata relevant to reviewers and maintainers.

Code Coverage & Benchmarks

For details see: https://corecheck.dev/bitcoin/bitcoin/pulls/33963.

Reviews

See the guideline for information on the review process. A summary of reviews will appear here.

maflcko commented at 8:24 pm on November 27, 2025: member

I don’t understand this change:

The tighter bounds check does not fix the false positive warning: https://godbolt.org/z/bn4sn5ehx
Using ranges fixes it, but likely only because they break the array-bounds analysis.
There is no CI task that runs into this warning
I can’t reproduce this on GCC 15.2 on Linux: https://github.com/maflcko/bitcoin-core-nightly/actions/runs/19735442483/job/56546112570#step:5:8
It reproduces with GCC 16, but there are many more warnings: https://github.com/maflcko/bitcoin-core-nightly/actions/runs/19735442483/job/56546112689#step:9:1188

No objection, but it is clear the compiler is at fault here, and fixing one warning and leaving the others around doesn’t seem that useful.

l0rinc closed this on Nov 27, 2025

l0rinc commented at 8:31 pm on November 27, 2025: contributor

Thanks for checking, it seems we have to be systematic about this.

l0rinc deleted the branch on Nov 27, 2025

test: tighten V2TransportTester::SendMessage bounds for GCC 15.2 #33963

Summary

Fix

Reproducer

Code Coverage & Benchmarks

Reviews

test: tighten `V2TransportTester::SendMessage` bounds for GCC 15.2 #33963