clang 3.2

[Ben Goodrich]

clang 3.2 is scheduled to be released December 16. So, if there are problems with it, we need to catch them quickly. Otherwise, they are more likely to delay the fix until 3.2.1 is released.

There are some binary RC2s here

http://llvm.org/pre-releases/3.2/rc2/

including an experimental one for Windows (actually, they have had experimental binaries for Windows since 3.1). If that works it would be great because there are probably some people that don't have enough RAM to compile some models on Windows.

For me, test-unit passes if all of the following are commented out:

From make/os_linux
ifeq (clang++,$(CC_TYPE))
#  CFLAGS += -std=c++0x
#  CFLAGS_GTEST += -DGTEST_USE_OWN_TR1_TUPLE
#  CFLAGS += -stdlib=libc++

From src/test/gm/parser_test.cpp
//  EXPECT_TRUE(is_parsable("src/test/gm/model_specs/function_signatures7.stan"));

If you uncomment  CFLAGS += -std=c++0x and CFLAGS_GTEST += -DGTEST_USE_OWN_TR1_TUPLE and use a version of boost  <= 1.51, then there are parser errors with O > 0. On one hand, sensitivity to the optimization level is a bit suspicious. On the other hand, you get the same behavior from clang 3.1.

Ben