[PATCH] D11374: Make DW_AT_[MIPS]linkage_name optional, and off by default for SCE.

Paul Robinson via llvm-commits llvm-commits at lists.llvm.org
Tue Aug 11 14:03:55 PDT 2015


probinson added a comment.

Thanks for the review!  Will commit after I've refreshed to current trunk (two weeks sees a lot of commits).


================
Comment at: test/DebugInfo/2010-04-06-NestedFnDbgInfo.ll:16
@@ -15,3 +15,3 @@
 ; CHECK-NOT: DW_TAG
-; CHECK: DW_AT_linkage_name {{.*}} "_ZZN1B2fnEvEN1A3fooEv"
+; CHECK: DW_AT_name {{.*}} "foo"
 ; And just double check that there's no out of line definition that references
----------------
echristo wrote:
> probinson wrote:
> > echristo wrote:
> > > Why this change? Probably want to change the RUN line if you want to make this clean to run on sce.
> > In this case it looked like the linkage-name was not inherently interesting, but just a way to validate that the CHECK stream was in the correct DIE.  By using the unmangled name instead, it preserves the purpose of the check while being debugger-feature-neutral.
> > I did something similar in cases like recursive_inlining.ll, where the mangled name was just a lazy way to verify that the cross-DIE references were pointing to the right places; verifying the actual references rather than the names seemed like an improvement on its own, as well as avoiding a dependence on having mangled names.
> > In other tests, the mangled name was either inherently part of the test, or not so easy to find a substitute for, and in those cases I added the command-line option.
> > 
> OK, as long as there's only a single foo I guess it makes sense. Seems a bit fragile which is what I was worried about.
There is nothing else named "foo" in this test.
We'd all be happier if the test had the original C++ source as commentary, but at least there is a comment saying that "foo" is a function within class "A" (see lines 5-6) so we don't have to worry about it being a ctor or anything like that.



http://reviews.llvm.org/D11374





More information about the llvm-commits mailing list