| David Blaikie | 683b0c5 | 2018-02-09 03:34:32 +0000 | [diff] [blame] | 1 | RUN: rm -rf %t && mkdir -p %t |
| 2 | RUN: cd %t |
| 3 | RUN: cp %p/Inputs/split-dwarf-no-skel-address.dwo %t |
| 4 | RUN: echo "%p/Inputs/split-dwarf-no-skel-address.o 0x4" > %t.input |
| 5 | |
| 6 | RUN: llvm-symbolizer --functions=linkage --inlining --demangle=false \ |
| 7 | RUN: --default-arch=i386 < %t.input | FileCheck %s |
| 8 | |
| 9 | Built from the following source: |
| 10 | void f1(); |
| 11 | __attribute__((always_inline)) inline void f2() { f1(); } |
| 12 | void f3() { f2(); } |
| 13 | |
| 14 | Compiled to assembly with clang, modified the skeleton CU to remove the |
| 15 | high/low pc (& update the CU length field and abbrev to match) & then |
| 16 | compile/objcopy to make the .o and .dwo. |
| 17 | |
| David Blaikie | 1e91943 | 2018-12-21 22:25:01 +0000 | [diff] [blame] | 18 | Ensure that the f2 inlined frame is not included - it's inefficient to have to |
| 19 | go and load all the debug info and search for the address ranges, so assume |
| 20 | that a lack of ranges on the CU means the CU covers no addresses. |
| 21 | |
| 22 | CHECK-NOT: _Z2f2v |