[PATCH] D63167: [Clang] Remove obsolete -enable-split-dwarf={single,split}

[Aaron Puchert via Phabricator via cfe-commits]

aaronpuchert added inline comments.


================
Comment at: include/clang/Basic/CodeGenOptions.def:263
 
-ENUM_CODEGENOPT(SplitDwarfMode, DwarfFissionKind, 2, NoFission) ///< DWARF fission mode to use.
+CODEGENOPT(EnableSplitDwarf, 1, 0) ///< Whether to enable split DWARF.
 
----------------
aaronpuchert wrote:
> dblaikie wrote:
> > Could we skip this and rely on "SplitDwarFile" being non-empty?
> I think not, have a look at test/CodeGen/split-debug-filename.c:
> 
> ```
> // RUN: %clang_cc1 -debug-info-kind=limited -split-dwarf-file foo.dwo -S -emit-llvm -o - %s | FileCheck %s
> // RUN: %clang_cc1 -debug-info-kind=limited -enable-split-dwarf -split-dwarf-file foo.dwo -S -emit-llvm -o - %s | FileCheck --check-prefix=VANILLA %s
> // ...
> 
> // Testing to ensure that the dwo name gets output into the compile unit.
> // CHECK: !DICompileUnit({{.*}}, splitDebugFilename: "foo.dwo"
> 
> // Testing to ensure that the dwo name is not output into the compile unit if
> // it's for vanilla split-dwarf rather than split-dwarf for implicit modules.
> // VANILLA-NOT: splitDebugFilename
> ```
> 
> This seems intentional.
Although if that's up to change, I think we can drop it. If I'm not overlooking anything, `-enable-split-dwarf` does nothing when `-emit-llvm` is also given. So I'm not sure we lose anything by treating it the same way as `-emit-llvm -split-dwarf-file ...` without `-enable-split-dwarf`.

Actually `llc` has exactly that behavior (inferring whether to enable Split DWARF depending on whether there was a `-split-dwarf-file` argument), as I remarked in D59673#1457720, but it doesn't emit IR and doesn't have to deal with implicit modules...


Repository:
  rC Clang

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D63167/new/

https://reviews.llvm.org/D63167