[Lldb-commits] [PATCH] D85376: Fix how ValueObjectChild handles bit-fields stored in a Scalar in UpdateValue()

[Adrian Prantl via Phabricator via lldb-commits]

aprantl added inline comments.


================
Comment at: lldb/test/API/functionalities/data-formatter/valueobj-pass-by-reg/Makefile:1
+EXE := a.out
+CFLAGS := -O1
----------------
I think this is redundant. The default is a.out


================
Comment at: lldb/test/API/functionalities/data-formatter/valueobj-pass-by-reg/Makefile:2
+EXE := a.out
+CFLAGS := -O1
+
----------------
davide wrote:
> davide wrote:
> > This is fundamentally a no-go. Depending on the optimization pipeline passes this in a register is an assumption that might go away at some point in the future and this test won't test what it has to & will still pass silently.
> > 
> > Something like this might work:
> > https://gcc.gnu.org/onlinedocs/gcc/Local-Register-Variables.html
> *depending on the optimization pipeline, the fact that is passed in a register is an assumption that
Given that the source code is a .s file, I think the -O1 is just redundant and can be removed. Using assembler is fine for this purpose.


================
Comment at: lldb/test/API/functionalities/data-formatter/valueobj-pass-by-reg/Makefile:8
+	$(CC) $(CFLAGS) $(SRCDIR)/main.s -c -o main.o
+	$(CC) $(CFLAGS) main.o -o a.out
----------------
Is it possible to just override the rule that produces the .o file? Otherwise you are dropping the codesign and dsymutil phase.


================
Comment at: lldb/test/API/functionalities/data-formatter/valueobj-pass-by-reg/TestValueObjPassByRef.py:15
+        self.runCmd("b f");
+        self.runCmd("run");
+
----------------
lldbutil has a helper for running to a breakpoint by name.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D85376/new/

https://reviews.llvm.org/D85376