On 5/31/24 12:38 PM, John Hubbard wrote:
...
diff --git a/tools/testing/selftests/x86/Makefile b/tools/testing/selftests/x86/Makefile
index d0bb32bd5538..5c8757a25998 100644
--- a/tools/testing/selftests/x86/Makefile
+++ b/tools/testing/selftests/x86/Makefile
@@ -40,6 +40,13 @@ CFLAGS := -O2 -g -std=gnu99 -pthread -Wall $(KHDR_INCLUDES)
# call32_from_64 in thunks.S uses absolute addresses.
ifeq ($(CAN_BUILD_WITH_NOPIE),1)
CFLAGS += -no-pie
+
+ifneq ($(LLVM),)
+# clang only wants to see -no-pie during linking. Here, we don't have a separate
+# linking stage, so a compiler warning is unavoidable without (wastefully)
+# restructuring the Makefile. Avoid this by simply disabling that warning.
+CFLAGS += -Wno-unused-command-line-argument
+endif
endif
define gen-target-rule-32
This actually can be improved slightly, as per our latest tentative
decision about how to handle both LLVM=1 and CC=clang cases [1].
If this series goes via Shuah's next tree, then I can put the
CC_IS_CLANG fix from [1] on top of previous patches, that will work
nicely.
In other words, I think we can use this series as-is, and let the
CC_IS_CLANG fix proceed just after that. Because it gets better a piece
at a time: first we get rid of the warning for most cases, then we get
rid of it for the odd "make CC=clang" case as well.
[1] https://lore.kernel.org/6b32399f-d9c6-4df5-b1e5-755ef4acf25d@xxxxxxxxxx
thanks,
--
John Hubbard
NVIDIA
