[PATCH] D71244: [DWARF5][DWARFVerifier] Check that Skeleton compilation unit does not have children.

Paul Robinson via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Dec 10 09:41:45 PST 2019


probinson added inline comments.


================
Comment at: clang/lib/Driver/ToolChains/Clang.cpp:3542
       Args.hasFlag(options::OPT_fsplit_dwarf_inlining,
-                   options::OPT_fno_split_dwarf_inlining, true);
+                   options::OPT_fno_split_dwarf_inlining, false);
 
----------------
aprantl wrote:
> What does this have to do with the dwarfdump patch?
So that by default we emit DWARF that passes the verifier?


================
Comment at: llvm/test/DebugInfo/skeleton-unit-verify.ll:3
+; RUN: llvm-dwarfdump --verify %t | FileCheck %s
+; RUN: sed 's/splitDebugInlining: false/splitDebugInlining: true/' < %s > %t2
+; RUN: llc %t2 -mtriple=x86_64-linux -filetype=obj -split-dwarf-cross-cu-references --split-dwarf-file=test.dwo -O3 -o %t1 
----------------
aprantl wrote:
> probinson wrote:
> > There is no llc option to control this?  (Sorry, I'm lazy, haven't looked...)
> You might want to use an assembler test instead. For broken debug info, that seems reasonable.
We want to be able to emit compliant DWARF (with splitDebugInlining = false) and DWARF that is useful for backtraces in the absence of the .dwo file, that we properly detect as non-compliant (with splitDebugInlining = true).


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71244/new/

https://reviews.llvm.org/D71244





More information about the llvm-commits mailing list