[PATCH] D26942: [SelectionDAG] getRawSubclassData should not return HasDebugValue.

[Chih-Hung Hsieh via Phabricator via llvm-commits]

chh added inline comments.


================
Comment at: test/CodeGen/X86/fp128-g.ll:1
+; RUN: llc < %s -O2 -mtriple=x86_64-linux-android -mattr=+mmx | FileCheck %s --check-prefix=X64
+
----------------
jlebar wrote:
> jlebar wrote:
> > jlebar wrote:
> > > Please add a comment in this file indicating what it's testing.  This is important so that someone without context to this thread who hits a failure in this file can understand what is going on.
> > This seems like a lot of testcase for this bug.  If all of these tests are necessary to cover cases here, please add comments explaining why they're different.
> Is the debug metadata necessary in this test?
The problem appeared only when compiled with -g.
I included all debug metadata generated from clang with the hope to catch any future regression related to -g.



================
Comment at: test/CodeGen/X86/fp128-g.ll:23
+; X64-LABEL: test_return1:
+; X64:       .loc
+; X64:       movaps     (%rdi), %xmm0
----------------
jlebar wrote:
> Do we care about the .loc directives in the output?
It's just a minimal check that debug info was not lost,
when we ignore debug bit in CSE node comparison
and legalization of fp128 types.
 


https://reviews.llvm.org/D26942