https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81733
--- Comment #11 from Iain Sandoe <iains at gcc dot gnu.org> --- Dominique identifies r249105 as triggering the issue (but this might well be exposing a latent). I have built xc 8.2.1 series tools on Darwin 10 and carried out the following test with trunk at 253320 (which bootstraps for me on Darwin15). 1. use the cctools assembler from cctools-895 and ld64-274.2 - this fails with the CFI issue 2. use an assembler based on LLVM-4.0.1 and ld64-274.2 - this succeeds. Thus, I suspect that the transition between "not working" and "working" Xcode versions corresponds to the switch in Xcode to default to using the LLVM based assembler. NOTE: "working" is quoted, since I haven't yet had a chance to analyse if the output is completely "correct" or just that the resultant output doesn't trigger the 0-length FDE assert.
