https://github.com/Michael137 created https://github.com/llvm/llvm-project/pull/161363
Starting with https://github.com/llvm/llvm-project/pull/148877 we started encoding the module ID of the function DIE we are currently parsing into its `AsmLabel` in the AST. When the JIT asks LLDB to resolve our special mangled name, we would locate the module and resolve the function/symbol we found in it. If we are debugging with a `SymbolFileDWARFDebugMap`, the module ID we encode is that of the `.o` file that is tracked by the debug-map. To resolve the address of the DIE in that `.o` file, we have to ask `SymbolFileDWARFDebugMap::LinkOSOAddress` to turn the address of the `.o` DIE into a real address in the linked executable. This will only work if the `.o` address was actually tracked by the debug-map. However, if the function definition appears in multiple `.o` files (which is the case for functions defined in headers), the linker will most likely de-deuplicate that definition. So most `.o`'s definition DIEs for that function won't have a contribution in the debug-map, and thus we fail to resolve the address. When debugging Clang on Darwin, e.g., you'd see: ``` (lldb) expr CXXDecl->getName() error: Couldn't look up symbols: $__lldb_func::0x1:0x4000d000002359da:_ZNK5clang9NamedDecl7getNameEv Hint: The expression tried to call a function that is not present in the target, perhaps because it was optimized out by the compiler. ``` unless you were stopped in the `.o` file whose definition of `getName` made it into the final executable. The fix here is to error out if we fail to resolve the address, causing us to fall back on the old flow which did a lookup by mangled name, which the `SymbolFileDWARFDebugMap` will handle correctly. An alternative fix to this would be to encode the `SymbolFileDWARFDebugMap`'s module-id. And implement `SymbolFileDWARFDebugMap::ResolveFunctionCallLabel` by doing a mangled name lookup. The proposed approach doesn't stop us from implementing that, so we could choose to do it in a follow-up. rdar://161393045 >From 313e603748a0a4f20835a6ebfc73afed18dfd956 Mon Sep 17 00:00:00 2001 From: Michael Buch <[email protected]> Date: Tue, 30 Sep 2025 13:34:14 +0100 Subject: [PATCH] [lldb][IRExecutionUnit] Return error on failure to resolve function address Starting with https://github.com/llvm/llvm-project/pull/148877 we started encoding the module ID of the function DIE we are currently parsing into its `AsmLabel` in the AST. When the JIT asks LLDB to resolve our special mangled name, we would locate the module and resolve the function/symbol we found in it. If we are debugging with a `SymbolFileDWARFDebugMap`, the module ID we encode is that of the `.o` file that is tracked by the debug-map. To resolve the address of the DIE in that `.o` file, we have to ask `SymbolFileDWARFDebugMap::LinkOSOAddress` to turn the address of the `.o` DIE into a real address in the linked executable. This will only work if the `.o` address was actually tracked by the debug-map. However, if the function definition appears in multiple `.o` files (which is the case for functions defined in headers), the linker will most likely de-deuplicate that definition. So most `.o`'s definition DIEs for that function won't have a contribution in the debug-map, and thus we fail to resolve the address. When debugging Clang on Darwin, e.g., you'd see: ``` (lldb) expr CXXDecl->getName() error: Couldn't look up symbols: $__lldb_func::0x1:0x4000d000002359da:_ZNK5clang9NamedDecl7getNameEv Hint: The expression tried to call a function that is not present in the target, perhaps because it was optimized out by the compiler. ``` unless you were stopped in the `.o` file whose definition of `getName` made it into the final executable. The fix here is to error out if we fail to resolve the address, causing us to fall back on the old flow which did a lookup by mangled name, which the `SymbolFileDWARFDebugMap` will handle correctly. An alternative fix to this would be to encode the `SymbolFileDWARFDebugMap`'s module-id. And implement `SymbolFileDWARFDebugMap::ResolveFunctionCallLabel` by doing a mangled name lookup. The proposed approach doesn't stop us from implementing that, so we could choose to do it in a follow-up. rdar://161393045 --- lldb/source/Expression/IRExecutionUnit.cpp | 7 ++++- .../function-call-from-object-file/Makefile | 3 ++ .../TestFunctionCallFromObjectFile.py | 29 +++++++++++++++++++ .../function-call-from-object-file/common.h | 8 +++++ .../function-call-from-object-file/lib1.cpp | 8 +++++ .../function-call-from-object-file/lib2.cpp | 6 ++++ .../function-call-from-object-file/main.cpp | 10 +++++++ 7 files changed, 70 insertions(+), 1 deletion(-) create mode 100644 lldb/test/API/lang/cpp/function-call-from-object-file/Makefile create mode 100644 lldb/test/API/lang/cpp/function-call-from-object-file/TestFunctionCallFromObjectFile.py create mode 100644 lldb/test/API/lang/cpp/function-call-from-object-file/common.h create mode 100644 lldb/test/API/lang/cpp/function-call-from-object-file/lib1.cpp create mode 100644 lldb/test/API/lang/cpp/function-call-from-object-file/lib2.cpp create mode 100644 lldb/test/API/lang/cpp/function-call-from-object-file/main.cpp diff --git a/lldb/source/Expression/IRExecutionUnit.cpp b/lldb/source/Expression/IRExecutionUnit.cpp index 25d4a87b89ef2..60b9de0d21b2e 100644 --- a/lldb/source/Expression/IRExecutionUnit.cpp +++ b/lldb/source/Expression/IRExecutionUnit.cpp @@ -751,7 +751,12 @@ ResolveFunctionCallLabel(FunctionCallLabel &label, sc_list.Append(*sc_or_err); LoadAddressResolver resolver(*sc.target_sp, symbol_was_missing_weak); - return resolver.Resolve(sc_list).value_or(LLDB_INVALID_ADDRESS); + lldb::addr_t resolved_addr = + resolver.Resolve(sc_list).value_or(LLDB_INVALID_ADDRESS); + if (resolved_addr == LLDB_INVALID_ADDRESS) + return llvm::createStringError("couldn't resolve address for function"); + + return resolved_addr; } lldb::addr_t diff --git a/lldb/test/API/lang/cpp/function-call-from-object-file/Makefile b/lldb/test/API/lang/cpp/function-call-from-object-file/Makefile new file mode 100644 index 0000000000000..285bbfbbca4fe --- /dev/null +++ b/lldb/test/API/lang/cpp/function-call-from-object-file/Makefile @@ -0,0 +1,3 @@ +CXX_SOURCES := main.cpp lib1.cpp lib2.cpp + +include Makefile.rules diff --git a/lldb/test/API/lang/cpp/function-call-from-object-file/TestFunctionCallFromObjectFile.py b/lldb/test/API/lang/cpp/function-call-from-object-file/TestFunctionCallFromObjectFile.py new file mode 100644 index 0000000000000..f0a7aef182a67 --- /dev/null +++ b/lldb/test/API/lang/cpp/function-call-from-object-file/TestFunctionCallFromObjectFile.py @@ -0,0 +1,29 @@ +""" +Tests that we can call functions that have definitions in multiple +CUs in the debug-info (which is the case for functions defined in headers). +The linker will most likely de-duplicate the functiond definitions when linking +the final executable. On Darwin, this will create a debug-map that LLDB will use +to fix up object file addresses to addresses in the linked executable. However, +if we parsed the DIE from the object file whose functiond definition got stripped +by the linker, LLDB needs to ensure it can still resolve the function symbol it +got for it. +""" + +import lldb +from lldbsuite.test.decorators import * +from lldbsuite.test.lldbtest import * +from lldbsuite.test import lldbutil + + +class TestFunctionCallFromObjectFile(TestBase): + def test_lib1(self): + self.build() + lldbutil.run_to_name_breakpoint(self, "lib1_func") + + self.expect_expr("Foo{}.foo()", result_type="int", result_value="15") + + def test_lib2(self): + self.build() + lldbutil.run_to_name_breakpoint(self, "lib2_func") + + self.expect_expr("Foo{}.foo()", result_type="int", result_value="15") diff --git a/lldb/test/API/lang/cpp/function-call-from-object-file/common.h b/lldb/test/API/lang/cpp/function-call-from-object-file/common.h new file mode 100644 index 0000000000000..76e23be6b97a6 --- /dev/null +++ b/lldb/test/API/lang/cpp/function-call-from-object-file/common.h @@ -0,0 +1,8 @@ +#ifndef COMMON_H_IN +#define COMMON_H_IN + +struct Foo { + int foo() { return 15; } +}; + +#endif // COMMON_H_IN diff --git a/lldb/test/API/lang/cpp/function-call-from-object-file/lib1.cpp b/lldb/test/API/lang/cpp/function-call-from-object-file/lib1.cpp new file mode 100644 index 0000000000000..b97bcc1b712b6 --- /dev/null +++ b/lldb/test/API/lang/cpp/function-call-from-object-file/lib1.cpp @@ -0,0 +1,8 @@ +#include "common.h" + +// Parameter "Foo*" forces LLDB to parse "Foo" from the object +// file that it is stopped in. +void lib1_func(Foo *) { + // Force definition into lib1.o debug-info. + Foo{}.foo(); +} diff --git a/lldb/test/API/lang/cpp/function-call-from-object-file/lib2.cpp b/lldb/test/API/lang/cpp/function-call-from-object-file/lib2.cpp new file mode 100644 index 0000000000000..2f9d81a8bdf4c --- /dev/null +++ b/lldb/test/API/lang/cpp/function-call-from-object-file/lib2.cpp @@ -0,0 +1,6 @@ +#include "common.h" + +void lib2_func(Foo *) { + // Force definition into lib2.o debug-info. + Foo{}.foo(); +} diff --git a/lldb/test/API/lang/cpp/function-call-from-object-file/main.cpp b/lldb/test/API/lang/cpp/function-call-from-object-file/main.cpp new file mode 100644 index 0000000000000..61ca798daf1df --- /dev/null +++ b/lldb/test/API/lang/cpp/function-call-from-object-file/main.cpp @@ -0,0 +1,10 @@ +struct Foo; + +extern void lib1_func(Foo *); +extern void lib2_func(Foo *); + +int main() { + lib1_func(nullptr); + lib2_func(nullptr); + return 0; +} _______________________________________________ lldb-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
