https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69004
--- Comment #18 from PeteVine <tulipawn at gmail dot com> --- Great, thanks! Now, as I mentioned in one of previous comments, the same t-engine build procedure works fine on x86, so we can be fairly certain I didn't break anything. The large amount of physfs/luajit profile data probably leads to some sort of corruption on ARM. I'm going to write down the steps after trying to reproduce this again myself. I've got gcc 6.1.1 and 7.0.0 from a week ago, apart from 4.9 and 5.3 - which version would you prefer to see used? (better debugging tips welcome)
