From mboxrd@z Thu Jan 1 00:00:00 1970 X-Msuck: nntp://news.gmane.org/gmane.linux.lib.musl.general/9316 Path: news.gmane.org!not-for-mail From: Rich Felker Newsgroups: gmane.linux.lib.musl.general Subject: Re: test results from latest git Date: Mon, 15 Feb 2016 15:11:22 -0500 Message-ID: <20160215201122.GU9349@brightrain.aerifal.cx> References: <20160215164511.GI14809@waldemar-brodkorb.de> <20160215192534.GR9915@port70.net> <20160215200540.GL14809@waldemar-brodkorb.de> Reply-To: musl@lists.openwall.com NNTP-Posting-Host: plane.gmane.org Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii X-Trace: ger.gmane.org 1455567106 14901 80.91.229.3 (15 Feb 2016 20:11:46 GMT) X-Complaints-To: usenet@ger.gmane.org NNTP-Posting-Date: Mon, 15 Feb 2016 20:11:46 +0000 (UTC) To: musl@lists.openwall.com Original-X-From: musl-return-9329-gllmg-musl=m.gmane.org@lists.openwall.com Mon Feb 15 21:11:39 2016 Return-path: Envelope-to: gllmg-musl@m.gmane.org Original-Received: from mother.openwall.net ([195.42.179.200]) by plane.gmane.org with smtp (Exim 4.69) (envelope-from ) id 1aVPUg-0004VB-Hd for gllmg-musl@m.gmane.org; Mon, 15 Feb 2016 21:11:38 +0100 Original-Received: (qmail 24185 invoked by uid 550); 15 Feb 2016 20:11:36 -0000 Mailing-List: contact musl-help@lists.openwall.com; run by ezmlm Precedence: bulk List-Post: List-Help: List-Unsubscribe: List-Subscribe: List-ID: Original-Received: (qmail 24164 invoked from network); 15 Feb 2016 20:11:36 -0000 Content-Disposition: inline In-Reply-To: <20160215200540.GL14809@waldemar-brodkorb.de> User-Agent: Mutt/1.5.21 (2010-09-15) Original-Sender: Rich Felker Xref: news.gmane.org gmane.linux.lib.musl.general:9316 Archived-At: On Mon, Feb 15, 2016 at 09:05:40PM +0100, Waldemar Brodkorb wrote: > Hi, > Szabolcs Nagy wrote, > > > * Waldemar Brodkorb [2016-02-15 17:45:11 +0100]: > > > > > > I uploaded latest test results from embedded-test here: > > > http://tests.embedded-test.org/musl/git-14022016/ > > > > > > > hm, almost all armv7 float/math tests failed > > > > can you give some info about the toolchain? (gcc -v) > > and the armv6 one for comparision? (which seems to work) > > ARMv6: > usr/bin/arm-openadk-linux-musleabihf-gcc -v > Using built-in specs. > COLLECT_GCC=./openadk/toolchain_qemu-arm-realview-eb-mpcore_musl_arm_hard_eabihf/usr/bin/arm-openadk-linux-musleabihf-gcc > COLLECT_LTO_WRAPPER=/home/wbx/embedded-test/openadk/toolchain_qemu-arm-realview-eb-mpcore_musl_arm_hard_eabihf/usr/libexec/gcc/arm-openadk-linux-musleabihf/5.3.0/lto-wrapper > Target: arm-openadk-linux-musleabihf > Configured with: > /home/wbx/embedded-test/openadk/toolchain_build_qemu-arm-realview-eb-mpcore_musl_arm_hard_eabihf/w-gcc-5.3.0-1/gcc-5.3.0/configure > --prefix=/home/wbx/embedded-test/openadk/toolchain_qemu-arm-realview-eb-mpcore_musl_arm_hard_eabihf/usr > --with-bugurl=http://www.openadk.org/ --build=x86_64-linux-gnu > --host=x86_64-linux-gnu --target=arm-openadk-linux-musleabihf > --with-gmp=/home/wbx/embedded-test/openadk/host_x86_64-linux-gnu/usr > --with-mpfr=/home/wbx/embedded-test/openadk/host_x86_64-linux-gnu/usr > --with-libelf=/home/wbx/embedded-test/openadk/host_x86_64-linux-gnu/usr > --disable-__cxa_atexit --with-system-zlib --with-gnu-ld > --with-gnu-as --disable-libsanitizer --disable-install-libiberty > --disable-libitm --disable-libmudflap --disable-libgomp > --disable-libcilkrts --disable-libquadmath > --disable-libquadmath-support --disable-decimal-float > --disable-libstdcxx-pch --disable-ppl-version-check > --disable-cloog-version-check --without-ppl --without-cloog > --without-isl --disable-nls --disable-lto --with-cpu=mpcore > --with-float=hard --enable-tls --enable-threads --enable-libatomic > --enable-shared=libstdc++,libgcc --disable-libssp --disable-biarch > --disable-multilib --enable-cxx-flags=-fPIC --enable-languages=c,c++ > --with-build-sysroot='/../../target_qemu-arm-realview-eb-mpcore_musl_arm_hard_eabihf' > --with-sysroot='/../../target_qemu-arm-realview-eb-mpcore_musl_arm_hard_eabihf' > Thread model: posix > gcc version 5.3.0 (GCC) > > ARMv7: > usr/bin/arm-openadk-linux-musleabihf-gcc -v > Using built-in specs. > COLLECT_GCC=./openadk/toolchain_qemu-arm-vexpress-a9_musl_arm_hard_eabihf/usr/bin/arm-openadk-linux-musleabihf-gcc > COLLECT_LTO_WRAPPER=/home/wbx/embedded-test/openadk/toolchain_qemu-arm-vexpress-a9_musl_arm_hard_eabihf/usr/libexec/gcc/arm-openadk-linux-musleabihf/5.3.0/lto-wrapper > Target: arm-openadk-linux-musleabihf > Configured with: > /home/wbx/embedded-test/openadk/toolchain_build_qemu-arm-vexpress-a9_musl_arm_hard_eabihf/w-gcc-5.3.0-1/gcc-5.3.0/configure > --prefix=/home/wbx/embedded-test/openadk/toolchain_qemu-arm-vexpress-a9_musl_arm_hard_eabihf/usr > --with-bugurl=http://www.openadk.org/ --build=x86_64-linux-gnu > --host=x86_64-linux-gnu --target=arm-openadk-linux-musleabihf > --with-gmp=/home/wbx/embedded-test/openadk/host_x86_64-linux-gnu/usr > --with-mpfr=/home/wbx/embedded-test/openadk/host_x86_64-linux-gnu/usr > --with-libelf=/home/wbx/embedded-test/openadk/host_x86_64-linux-gnu/usr > --disable-__cxa_atexit --with-system-zlib --with-gnu-ld > --with-gnu-as --disable-libsanitizer --disable-install-libiberty > --disable-libitm --disable-libmudflap --disable-libgomp > --disable-libcilkrts --disable-libquadmath > --disable-libquadmath-support --disable-decimal-float > --disable-libstdcxx-pch --disable-ppl-version-check > --disable-cloog-version-check --without-ppl --without-cloog > --without-isl --disable-nls --disable-lto --with-cpu=cortex-a9 > --with-float=hard --with-fpu=neon-fp16 --enable-tls --enable-threads > --enable-libatomic --enable-shared=libstdc++,libgcc --disable-libssp > --disable-biarch --disable-multilib --enable-cxx-flags=-fPIC > --enable-languages=c,c++ > --with-build-sysroot='/../../target_qemu-arm-vexpress-a9_musl_arm_hard_eabihf' > --with-sysroot='/../../target_qemu-arm-vexpress-a9_musl_arm_hard_eabihf' > Thread model: posix > gcc version 5.3.0 (GCC) > > I do not configure any --with-fpu for ARMv6. Are NEON intrinsics the > problem? OK, Neon is almost certainly your problem. It should never be used as the default for a toolchain because it produces non-IEEE-conforming results. You can enable Neon for individual applications where you don't care (just like -ffast-math) but using it to build libc will result in a broken libc, and using it by default could break other apps too. Perhaps we should add something to musl's configure to detect this misconfiguration and error out or disable Neon? Rich