Hi!
On Mon, 28 Oct 2024 at 08:56, Li, Pan2 via Gcc-regression gcc-regression@gcc.gnu.org wrote:
I have a try with below command but get error like "cc1: error: '-mfloat-abi=hard': selected architecture lacks an FPU". Linux ubuntu-arm 4.15.0-20-generic #21-Ubuntu SMP Tue Apr 24 06:16:20 UTC 2018 aarch64 aarch64 aarch64 GNU/Linux
_INSTALL_DIR=$1 ../configure --target=arm-linux-gnueabihf \ --enable-language=c,c++ \ --enable-bootstrap \ --disable-multilib \ --with-arch=armv7-a \ --prefix=${_INSTALL_DIR} make -j64 && make install
Any suggestions for reproducing this issue ?
It seems the problem is even more tricky than "just" bootstrap: AFAICS the "bootstrap" builds passed, the "bootstrap_profiled" and "bootstrap_profiled_lto_lean" failed.
So, you need to configure with --enable-bootstrap and build with 'make profiledbootstrap'.
In addition, the problem occurs on arm targets, as opposed to aarch64, so the easiest for you is probably to start a container in aarch32 (aka arm) mode on an aarch64 machine (provided its CPU supports that mode). Once in the container, you don't need to use --target XXX --with-arch=YYY configure options.
Good luck,
Christophe
Pan
-----Original Message----- From: Li, Pan2 Sent: Sunday, October 27, 2024 3:41 PM To: Sam James sam@gentoo.org Cc: gcc-regression@gcc.gnu.org; jeffreyalaw@gmail.com; juzhe.zhong@rivai.ai; kito.cheng@gmail.com; linaro-toolchain@lists.linaro.org; rdapp@ventanamicro.com Subject: RE: [Linaro-TCWG-CI] gcc-15-4672-gdf4af89bc3e: Failure on arm
I see, this error may require bootstrap build. There is no aarch64 machine currently. Let me try to find one somewhere.
Pan
-----Original Message----- From: Sam James sam@gentoo.org Sent: Sunday, October 27, 2024 3:37 PM To: Li, Pan2 pan2.li@intel.com Cc: gcc-regression@gcc.gnu.org; jeffreyalaw@gmail.com; juzhe.zhong@rivai.ai; kito.cheng@gmail.com; linaro-toolchain@lists.linaro.org; rdapp@ventanamicro.com Subject: RE: [Linaro-TCWG-CI] gcc-15-4672-gdf4af89bc3e: Failure on arm
It's not possible to do a profiledbootstrap via cross. It might be possible for the Linaro CI to extract the bad preprocessed source and share the options used to build it (options at least should be in the log) but that won't help if the compiler is miscompiled...