Comment 9 for bug 1246802

I changed two instances of __last - 2 to __last - 1. I'm running against r203836 on gcc-4_8-branch.

The thing that caught it was an experiment, not part of my test suite, so this is a little tricky. It already survived longer than the unpatched version did. The crash wasn't the same segfault, though -- it was a much later crash due to memory corruption. I'll let it run for a while under valgrind to see what, if anything, pops up.