When/why does xcode at the C-linker flag -no_deduplicate to C++ MacOS projects?

Hello all,

I'm still having this problem -- any help??

stp

We got the additional information that it only happens when building the debug configuration; building the release of the app links fine.

HELP!

stp

Hi stpope,

The flag -no_deduplicate is not a clang flag but an OSX ld flag.

If you type man ld on a terminal, you will see the list of parameters and get something like:

     -no_deduplicate
                 Don't run deduplication pass in linker

Normally this flag has an -Xlinker just before. This indicate that the no_duplicate flag is not for the clang driver but for the linker. I read somewhere on the Net that this is a Xcode bug, so maybe you need to update your installation or play with the Linking section in your Build Settings project on Xcode.

Hope this help.

Johan

I believe the -O0 flag (which is common in Debug configurations) winds up trigerring the -no_deduplicate flag. In Xcode, you can check the corresponding settings for optimization. If disabled (-O0), you may get -no_deduplicate.

I encontered the same issue when building from command line using Make. My Makefile sets:

CC=g++

to ensure everything is built with the C++ compiler (ultimately Clang -x c++). g++ is a 'compiler driver' that, in turn, calls the actual compiler or linker. These compiler drivers can translate the command-line flags in unexpected ways.

To resolve this, I revised the make rule I use to build the dylib so it does not pass CFLAGS or CXXFLAGS to the compiler driver. This avoids passing -O0 to the compiler driver. Instead, I pass only LDFLAGS and TARGET_ARCH to the compiler driver. The make rule or dylib I use is based on Make implicit rules for building an executable:

$(CC) $(LDFLAGS) $(TARGET_ARCH) $^ $(LDLIBS) -0 $@

Note that to build a dylib requires:

LDFLAGS=-dynamiclib