https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116686
--- Comment #8 from Edwin Lu <ewlu at rivosinc dot com> --- (In reply to Robin Dapp from comment #7) > Hmm, I don't fully understand. We're actually building with zvl256b right, > zvl1024b is first and thus gets overridden? But with zvl256b and QEMU > vlen=256 I'm not seeing a FAIL either. > > Edwin, can you try reproducing this on the current trunk? Looking at our postcommit logs, the test appears to be rather flakey so maybe the bisection is off https://github.com/patrick-rivos/gcc-postcommit-ci/issues?q=gcc.target%2Friscv%2Frvv%2Fautovec%2Fpr114734.c%20label%3Arv64 We perform each build in separate environments so even though we build zvl(128|256|512|1024)b, this one should be configured using the riscv-gnu-toolchain with configure --prefix=$(build_dir) --with-multilib-generator="rv64gcv_zvl1024b-lp64d--:-mrvv-max-lmul=m2 I think this builds the zvl1024b glibc but tries running it in zvl256b land which is causing a mismatch?
