https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81101
--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> --- (In reply to David Malcolm from comment #1) > End-users aren't meant to see this; this is for people building GCC from > source, with checking enabled (the self-tests are a no-op in a release > build). > > So the "user" in question here is a developer, packager, or (very) early > adopter. > > We want a backtrace when a selftest fails, and for it to abort the process. But is failing because the user gave a bogus path a test failure, or user error? I was trying to reproduce an error a user reported on irc, and tried using the -fself-test=. arg with a dummy path, and the output looked like a crash, rather than a self-test failure. If you think the distinction doesn't matter feel free to close this.