You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
ci: linux64_asan fails on Qt QDBusConnectionManager leak #7561
The new linux64_asan job fails in qt/test/test_dash-qt after all Qt tests pass because LeakSanitizer reports four small allocations from Qt's QDBusConnectionManager::executeConnectionRequest:
In both jobs, every Qt test reports success before LeakSanitizer terminates the test binary.
Root cause
This is a dependency-level Qt DBus shutdown leak, not a test assertion or Dash wallet failure. The leak stack is entirely in Qt/DBus connection management, and the same signature and byte count occur on unmodified develop.
Reproduction
Check out e25f9925df4f0bafaf5f6d13ac34e14bdc8eecc9.
Run the repository's linux64_asan CI target (the configuration in ci/test/00_setup_env_native_asan.sh, with leak detection enabled and test/sanitizer_suppressions/lsan).
Observe that qt/test/test_dash-qt passes its test cases, then exits 1 with the four QDBusConnectionManager::executeConnectionRequest allocations above.
Suggested direction
Add a narrowly scoped LSAN suppression for this Qt dependency stack (or ensure the Qt DBus connection manager is shut down before process exit), then verify no Dash-owned leaks are hidden.
Both jobs pass all Qt test cases and then report the same QDBusConnectionManager::executeConnectionRequest signature: 86 bytes in 4 allocations. This confirms the dependency-level Qt/DBus LSAN failure remains present on current develop and is not caused by #7479's wallet-only diff.
Both runs pass every Qt test and then fail at shutdown with the same QDBusConnectionManager::executeConnectionRequest signature: 86 bytes in 4 allocations. PR #7483 changes net processing, LLMQ signing, and non-Qt test support; it does not touch Qt/DBus. This matches the identical current-develop failure already documented here and confirms the PR failure is unrelated to its diff.
Activity
thepastaclaw commentedon Aug 8, 2026
Another identical occurrence:
66d72b4aa17cc4329512054d3dd7016c47c46e2d: run 31269700511 / linux64_asan job 93134790520develophead5c44379d739438ef71f8b64cbef8c8b598eed563: run 31271294064 / linux64_asan job 93140878081Both jobs pass all Qt test cases and then report the same
QDBusConnectionManager::executeConnectionRequestsignature: 86 bytes in 4 allocations. This confirms the dependency-level Qt/DBus LSAN failure remains present on currentdevelopand is not caused by #7479's wallet-only diff.thepastaclaw commentedon Aug 9, 2026
Another confirmed occurrence on PR #7483:
43cefa64ed6122e553de1c2a4e8105feb735208d: run 31271062102 / linux64_asan job 931403878366a2c939e46014c8bad9940fce1d9db95be2764a8: run 31277137627 / linux64_asan job 93152614501Both runs pass every Qt test and then fail at shutdown with the same
QDBusConnectionManager::executeConnectionRequestsignature: 86 bytes in 4 allocations. PR #7483 changes net processing, LLMQ signing, and non-Qt test support; it does not touch Qt/DBus. This matches the identical current-developfailure already documented here and confirms the PR failure is unrelated to its diff.