probable use-after-free errors

[PHHargrove]

I have a nightly tester running clang-upc (w/o upcr) against the Berkeley test suite on Mac OS X 10.8 once per week. The results this morning show the following two compiler crashes:

cd /usr/local/upc/testers64/cupc/runtime/work/dbg/upc-tests/bugzilla
/usr/local/upc/llvm-upc/bin/clang-upc -Werror=pointer-arith -g    -o testenv testenv.upc
dsymutil(80446,0x7fff7e383180) malloc: *** error for object 0x7fe669c07140: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug
clang-3.4: error: unable to execute command: Abort trap: 6
clang-3.4: error: dsymutil command failed due to signal (use -v to see invocation)

cd /usr/local/upc/testers64/cupc/runtime/work/dbg/upc-tests/mupc
/usr/local/upc/llvm-upc/bin/clang-upc -Werror=pointer-arith -g   -DDATA=long -o test_stress_04-long test_stress_04.c
dsymutil(88904,0x7fff7e383180) malloc: *** error for object 0x7fa69bc07180: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug
clang-3.4: error: unable to execute command: Abort trap: 6
clang-3.4: error: dsymutil command failed due to signal (use -v to see invocation)

Note that both errors are for compilations which were passed -g and that the same tests compiled with -O did not crash. Additionally, clang-upc2c was tested today from from the same build, and did not crash translating either of these tests.

Finally, it is worth noting that when this test configuration last ran on June 14 neither crash was present.

[PHHargrove]

Attempts to reproduce manually do not crash, despite using the same build and same command lines. Sigh.