https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71892
--- Comment #2 from Kern Sibbald <kern at sibbald dot com> --- Yes, we are aware of the option and how to fix the problem. The issue is that this optimization at low levels of -O1 and -O2 is not reasonable, and it is unreasonable and irresponsible to make such changes. Just search Internet to see what kinds of pains this creates -- all for nothing. Some years ago, we just accepted these kinds of crippling and costly changes. This incident has caused bad versions of Bacula to be distributed by at least two popular distros that seg fault because of g++. We are not the only project to be affected. For me this comes down to the philosophy of how the gcc project treats its users. Do you force risky changes on your users or do you try to protect them. The gcc project has its view, but this result is not acceptable to me. Some years ago there was no alternative to g++, but faced with these kind of problems that take months to fix because of an "unstable" and "incompatible" new compiler releases, I for one will now take a careful look at the alternatives. I suggest that you (the gcc project) carefully reconsider whether making such assumptions leading to risky optimizations is best practice.
