In function ‘gen_helper_get_dr’, inlined from ‘disas_insn’ at ../target/i386/tcg/translate.c:8174:17: /var/tmp/portage/app-emulation/qemu-6.2.0-r2/work/qemu-6.2.0/include/qemu/compiler.h:176:35: error: call to ‘qemu _build_not_reached_always’ declared with attribute error: code path is reachable 176 | #define qemu_build_not_reached() qemu_build_not_reached_always() | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ../target/i386/tcg/translate.c:192:7: note: in expansion of macro ‘qemu_build_not_reached’ 192 | { qemu_build_not_reached(); } | ^~~~~~~~~~~~~~~~~~~~~~ ../target/i386/tcg/translate.c:208:1: note: in expansion of macro ‘STUB_HELPER’ 208 | STUB_HELPER(get_dr, TCGv ret, TCGv_env env, TCGv_i32 reg) | ^~~~~~~~~~~
Created attachment 761669 [details] build log
Are you able to test if this is ppc64-specific?
This might be to do with, for some reason, building TCG when it.. shouldn't be? It's meant to be disabled right now but configure at least shows it being enabled for some parts.
(In reply to John Helmert III from comment #2) > Are you able to test if this is ppc64-specific? Same ebuild on amd64 builds fine. Only difference was USE=io-uring, which I don't think should be relevant.
Same problem here on x86_64 platform with qemu-6.2.0 please find info in attachments i also tried with MAKEOPTS="" - but get same error thanks, marcus.
Created attachment 762152 [details] build.log qemu-6.2.0_x86_64
Created attachment 762153 [details] emerge --info '=app-emulation/qemu-6.2.0::gentoo'
Created attachment 762154 [details] emerge -pqv '=app-emulation/qemu-6.2.0::gentoo'
I compared output of compiler commandline when compiling qemu-6.2.0 by hand (with identical call to configure) vs. compiling using portage: when using portage i have -O1 and -O2 flags and compilation of this particular file fails, when compiling by hand, i only have -O2 and compilation works. so i removed -O1 from /etc/portage/make.conf (CFLAGS) and now it compiles just fine! so the question remains where this -O2 comes from ? and i wonder how -O1 can influence -O2 ... but anyhow ... maybe someone can add a check in the ebuild to be clear on that fact - or to remove -O1 if it's set ... ?
