[llvm] [Symbolizer][Wasm] Use wasm-specific getSymbolSize to compute symbol sizes (PR #82083)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-binary-utilities

Author: Derek Schuff (dschuff)

<details>
<summary>Changes</summary>

getSymbolSize was recently added to WasmObjectFile and has correct sizes for
most symbol types. This makes llvm-symbolizer correctly symbolize addresses
in the middle of the symbol.

When reworking the test I also noticed that the DWARF info seems to be wrong
for the first instruction in each function. I noted that in the test comments
but didn't attempt to fix here.

---
Full diff: https://github.com/llvm/llvm-project/pull/82083.diff


2 Files Affected:

- (modified) llvm/lib/Object/SymbolSize.cpp (+7) 
- (modified) llvm/test/tools/llvm-symbolizer/wasm-basic.s (+44-9) 


``````````diff
diff --git a/llvm/lib/Object/SymbolSize.cpp b/llvm/lib/Object/SymbolSize.cpp
index cb20feffb710bb..635cd8373afbfb 100644
--- a/llvm/lib/Object/SymbolSize.cpp
+++ b/llvm/lib/Object/SymbolSize.cpp
@@ -65,6 +65,13 @@ llvm::object::computeSymbolSizes(const ObjectFile &O) {
     return Ret;
   }
 
+  if (const auto *E = dyn_cast<WasmObjectFile>(&O)) {
+    for (SymbolRef Sym : E->symbols()) {
+      Ret.push_back({Sym, E->getSymbolSize(Sym)});
+    }
+    return Ret;
+  }
+
   // Collect sorted symbol addresses. Include dummy addresses for the end
   // of each section.
   std::vector<SymEntry> Addresses;
diff --git a/llvm/test/tools/llvm-symbolizer/wasm-basic.s b/llvm/test/tools/llvm-symbolizer/wasm-basic.s
index cc189abcfca80c..1f425e52593167 100644
--- a/llvm/test/tools/llvm-symbolizer/wasm-basic.s
+++ b/llvm/test/tools/llvm-symbolizer/wasm-basic.s
@@ -1,24 +1,59 @@
 # REQUIRES: webassembly-registered-target
 # RUN: llvm-mc -triple=wasm32-unknown-unknown -filetype=obj %s -o %t.o -g
+# RUN: llvm-symbolizer --basenames --output-style=GNU -e %t.o 1 2 3 4 5 6 7 8 9 10 11 12 13 | FileCheck %s
 
 foo:
     .functype foo () -> ()
     nop
+    return
     end_function
 
 bar:
     .functype bar (i32) -> (i32)
     local.get 0
+    nop
     return
     end_function
 
-# RUN: llvm-symbolizer -e %t.o 3 4 7 8 | FileCheck %s
-## Byte 1 is the function length and 2 is the locals declaration.
-## Currently no line corresponds to them.
-## TODO: create a loc for .functype?
 
-## Test 2 functions to ensure wasm's function-sections system works.
-# CHECK: wasm-basic.s:6:0
-# CHECK: wasm-basic.s:7:0
-# CHECK: wasm-basic.s:11:0
-# CHECK: wasm-basic.s:11:0
+## Symbols start from (including) the function length and should cover all the
+## way to the next symbol start.
+## TODO: create a loc for .functype? It could go with the local declarations.
+
+## Byte 1 is the function length, has no loc but the symbol table considers it
+## the start of the function
+# CHECK: foo
+# CHECK-NEXT: ??:0
+## Byte 2 is the local declaration, but for some reason DWARF is marking it as line 7.
+## TODO: figure out why.
+# CHECK-NEXT: foo
+# CHECK-NEXT: wasm-basic.s:7
+## Byte 3 is actually the nop, line 7
+# CHECK-NEXT: foo
+# CHECK-NEXT: wasm-basic.s:7
+## Byte 4 is the return, line 8
+# CHECK-NEXT: foo
+# CHECK-NEXT: wasm-basic.s:8
+## Byte 5 is the end_function, line 9
+# CHECK-NEXT: foo
+# CHECK-NEXT: wasm-basic.s:9
+## Byte 6 is bar's function length, symbol table considers it part of bar
+# CHECK-NEXT: bar
+# CHECK-NEXT: ??:0
+## Byte 7 bar's local declaration, but DWARF marks it as line 13, like above
+# CHECK-NEXT: bar
+# CHECK-NEXT: wasm-basic.s:13
+## Byte 8 and 9 are actually the local.get on line 13
+# CHECK-NEXT: bar
+# CHECK-NEXT: wasm-basic.s:13
+# CHECK-NEXT: bar
+# CHECK-NEXT: wasm-basic.s:13
+## Byte 10 is the nop
+# CHECK-NEXT: bar
+# CHECK-NEXT: wasm-basic.s:14
+## Byte b is the return
+# CHECK-NEXT: bar
+# CHECK-NEXT: wasm-basic.s:15
+## Byte c is end_function
+# CHECK-NEXT: bar
+# CHECK-NEXT: wasm-basic.s:16

``````````

</details>


https://github.com/llvm/llvm-project/pull/82083