Support for older OS X versions

[mbudisic]

Sometimes between now and Snow Leopard, Apple made a change in their OS X standard libraries. This means that binaries compiled in Mavericks and later might not work on Snow Leopard and earlier. An example error generated is
Symbol not found: __sincos_stret
...
Expected in /usr/lib/libSystem.B.dylib.
The fix should be to issue -mmacosx-version-min=10.8 (or earlier) during makedist compile step in braidlab.

See StackExchange:
http://stackoverflow.com/questions/19015780/sincos-stret-undefined-symbol-when-linking here
http://stackoverflow.com/questions/2923502/what-does-macosx-version-min-imply

I'll see if this does the trick.

[mbudisic]

mmacosx-version-min=10.8 forces C++ compiler to link against libstdc++ over (the default for > 10.8) libc++. This is exactly what it should be, but now linking fails because somehow wrong symbols are being defined. For example

Building with 'Xcode Clang++'.
Undefined symbols for architecture x86_64:
  "std::ios_base::Init::Init()", referenced from:
      __GLOBAL__I_a in randomwalk_helper.o
  "std::ios_base::Init::~Init()", referenced from:
      __GLOBAL__I_a in randomwalk_helper.o
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)

This happens whenever one of the headers from stdlib is used (iostream, vector, etc...)

[mbudisic]

This is a pretty painful thing to hunt down... do we have someone running Snow Leopard? :)

[jeanluct]

Is this in response to a user?

[mbudisic]

Yes, Cecily at UCSD. I'm trying to see if she can compile on her end.

It looks like most of the things still run fine except, surprisingly, randomwalk_helper.cpp. Now, if I define MACOSX_DEPLOYMENT_TARGET=10.8 which should target her version of the OS, I cannot link things on my end. But this is the way that Apple approaches issue of backward compatibility so it's hard to say what is going wrong...

[jeanluct]

Has she considered upgrading OSX? It's a free upgrade... Snow Leopard is
getting long in the tooth, so to speak.

On Fri, Dec 12, 2014 at 5:09 PM, Marko Budisic notifications@github.com
wrote:

Yes, Cecily at UCSD. I'm trying to see if she can compile on her end.

It looks like most of the things still run fine except, surprisingly,
randomwalk_helper.cpp. Now, if I define MACOSX_DEPLOYMENT_TARGET=10.8
which should target her version of the OS, I cannot link things on my end.
But this is the way that Apple approaches issue of backward compatibility
so it's hard to say what is going wrong...

—
Reply to this email directly or view it on GitHub
#102 (comment).

[mbudisic]

Yeah, that's definitely true. No pun intended? :)

[mbudisic]

I'm putting this on hold until I reconnect with Cecily as she's the only one who had an interest in this.

[jeanluct]

Ok, I'll close it. Reopen if necessary.