Hi Luc,

It was strange, because it only appeared after I updated our mac-mini nightly build machine with the latest OS X and Xcode updates the other day. However mine and Graeme's mac book pros both have the latest Apple updates installed and it appears to work fine on our machines (although I'm not certain either of us actually tried a full build from scratch running the bootstrap.py script).

Cheers,

Richard

$ clang --version
Apple LLVM version 6.1.0 (clang-602.0.49) (based on LLVM 3.6.0svn)
Target: x86_64-apple-darwin14.3.0
Thread model: posix

$ sw_vers -productVersion
10.10.3

Xcode must be latest version (whatever that is).


On 14 April 2015 at 08:22, Luc Bourhis <luc_j_bourhis@mac.com> wrote:

On 14 Apr 2015, at 08:44, Richard Gildea <rgildea@gmail.com> wrote:

I managed to 'fix' the error with this commit:


However, I have no idea why the previous code was problematic. Adding a SCITBX_ASSERT(vi != 0); in the 'else' statement also made the error go away. I can have another look and try out your suggestions later today to see if that results in any more insight into the underlying problem.

My suggestions only aimed at zeroing on the problem. Now that you have found a fix, you won't learn more by following my suggestions. It looks like a bug in the optimiser of whatever version of clang/llvm Apple uses in the version of Xcode installed on that machine. Note clang optimiser had troubles with that file a couple of  years ago: it was missing some optimisations and the produced code was too slow. So now it looks too eager! So classic unfortunately… 

For the record, could you paste the outcome of "clang --version" on that machine? And tell us the version of Xcode as well.

I will move to Yosemite soon-ish but for the time being, I can't help much.

Best wishes,

Luc


_______________________________________________
cctbxbb mailing list
cctbxbb@phenix-online.org
http://phenix-online.org/mailman/listinfo/cctbxbb