[Lldb-commits] [PATCH] D71487: [LLDB] Fix address computation for inline function

Tue Dec 17 01:26:18 PST 2019

labath added a comment.

Thanks. I think we're getting closer, and I think we can start talking about a more direct way to test this behavior.

This is now definitely, as Adrian called it, a DWARF peculiarity, and so an assembly file with hard-coded dwarf would be much better. Can you take a look at the files in `test/Shell/SymbolFile/DWARF` and create something similar. It don't think you'll need very complicated dwarf -- it should be sufficient to have two compile units with one function each, and set one of the function's DW_AT_low_pc to zero. The test command could be something like running `lldb-test symbols %t -find=function -name=<foo>` and checking that it returns just a single entry...

================
Comment at: lldb/source/Plugins/SymbolFile/DWARF/SymbolFileDWARF.cpp:2358-2360
+    if (die.GetAttributeValueAsAddress(DW_AT_low_pc, /*fail_value=*/1) == 0 &&
+        !GetObjectFile()->GetModule()->HaveSectionAtAddressZero())
+      continue;
----------------
I think this check should go into the `ResolveFunction` call, as that function already does some validation of the function address. 

And I still believe that we shouldn't special case address zero, but rather just reject any address that does not map onto a known section (it should be sufficient to replace the `addr.IsValid()` check in `ResolveFunction` with `addr.IsSectionOffset()`). Functions not mapping to any sections are going to be mostly useless anyway, because the first thing that lldb does when resolving an address, is to resolve the module it should search in. It uses the section list to do that, and so if the function does not belong to any section, lldb will never find it, as it will not even get around to asking the Module about it.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71487/new/

https://reviews.llvm.org/D71487