Disable MSan origin tracking on the SystemLog flush hot path

Algunenano · Algunenano · commit 7017c20fd699 · 2026-05-14T19:58:53.000+02:00
Even at track-origins=1, the per-call origin maintenance dominates two hot helpers used by every QueryLog flush: - ProfileEvents::dumpToMapColumn (Interpreters/ProfileEventsExt.cpp) - SettingsImpl::dumpToMapColumn (Core/Settings.cpp, already covered) Each walks hundreds of keys and does tight LowCardinality::insertData calls in a loop, so the per-call __msan_set_alloca_origin overhead adds up to seconds per `SYSTEM FLUSH LOGS query_log` and the failing tests on the MSan WasmEdge 2/2 shard hit the 180 s flush-wait timeout. The CI logs show Block phase up to 14.7 s for a 5759-entry batch, dominated by these two functions. Extend the existing per-file override added in ClickHouse#92716 to also cover ProfileEventsExt.cpp and SystemLog.cpp. These files are leaf helpers where MSan checks don't carry uninit values into user code, so turning origin tracking off here is safe; the rest of the build keeps origin tracking.
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
@@ -231,11 +231,24 @@ else()
     target_compile_definitions(clickhouse_common_io PUBLIC ENABLE_DISTRIBUTED_CACHE=0)
 endif()
 
-# Settings dumpToMap is too slow with MSAN
+# Origin tracking dominates two flush hot paths under MSan: Settings::dumpToMapColumn (in
+# Core/Settings.cpp) and ProfileEvents::dumpToMapColumn (in Interpreters/ProfileEventsExt.cpp).
+# Both walk hundreds of keys and do tight LowCardinality::insertData calls in a loop, so the
+# per-call __msan_set_alloca_origin / __msan_chain_origin cost adds up to seconds per
+# `SYSTEM FLUSH LOGS query_log` (test 01548_query_log_query_execution_ms etc. were timing out
+# on the MSan WasmEdge stateless check after x86-64-v3 became the default). SystemLog.cpp is
+# included because `flushImpl` is on the same thread and contributes to the same bottleneck.
+# Origin tracking is still on globally for the rest of the build; these files are leaf
+# helpers where MSan checks don't carry uninit values into user code, so the trade is safe.
+# History: Core/Settings.cpp got this treatment in #92716; the other two were added when the
+# same pattern surfaced on QueryLog flush under v3.
+set_source_files_properties(Core/Settings.cpp                  PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer")
+set_source_files_properties(Interpreters/ProfileEventsExt.cpp  PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer")
+set_source_files_properties(Interpreters/SystemLog.cpp         PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer")
 if (SANITIZE STREQUAL "memory")
-    set_source_files_properties(Core/Settings.cpp PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer -fno-sanitize-memory-track-origins")
-else ()
-    set_source_files_properties(Core/Settings.cpp PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer")
+    set_source_files_properties(Core/Settings.cpp                  PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer -fno-sanitize-memory-track-origins")
+    set_source_files_properties(Interpreters/ProfileEventsExt.cpp  PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer -fno-sanitize-memory-track-origins")
+    set_source_files_properties(Interpreters/SystemLog.cpp         PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer -fno-sanitize-memory-track-origins")
 endif ()
 
 set_source_files_properties(Common/ThreadFuzzer.cpp PROPERTIES COMPILE_FLAGS "-fomit-frame-pointer -momit-leaf-frame-pointer")
