Source: ruby-build Version: 20241225.2-1 Severity: serious Hi,
See https://ci.debian.net/packages/r/ruby-build/unstable/armhf/ The error message in the build log is: cc -I. -Iinclude -Iapps/include -fPIC -pthread -march=armv7-a -Wa,--noexecstack -Wall -O3 -DOPENSSL_USE_NODELETE -DOPENSSL_PIC -DOPENSSLDIR="\"/tmp/autopkgtest.TtHp4o/autopkgtest_tmp/openssl/ssl\"" -DENGINESDIR="\"/tmp/autopkgtest.TtHp4o/autopkgtest_tmp/openssl/lib/engines-3\"" -DMODULESDIR="\"/tmp/autopkgtest.TtHp4o/autopkgtest_tmp/openssl/lib/ossl-modules\"" -DOPENSSL_BUILDING_OPENSSL -DZLIB -DZLIB_SHARED -DNDEBUG -MMD -MF apps/lib/libapps-lib-app_rand.d.tmp -MT apps/lib/libapps-lib-app_rand.o -c -o apps/lib/libapps-lib-app_rand.o apps/lib/app_rand.c cc1: error: '-mfloat-abi=hard': selected architecture lacks an FPU cc1: error: '-mfloat-abi=hard': selected architecture lacks an FPU cc1: error: '-mfloat-abi=hard': selected architecture lacks an FPU make[1]: *** [Makefile:4291: apps/lib/libapps-lib-app_params.o] Error 1 I think that what happens is: * the 'build-ruby-openssl' test builds the ruby interpreter with a vendored openssl (not the system one), so first it builds openssl (3.0.15) * openssl's configure results in explicitely using -march=armv7-a (rather than using the compiler's default, which would be -march=armv7-a+fp) * but the code being compiled has FP instructions * so GCC refuses to compile that code Those links sound relevant: 1/ https://github.com/checkpoint-restore/criu/issues/1653 > Starting with gcc-11, Debian's armhf compiler no longer builds with a > default -mfpu= option. Instead it enables the FPU via an extension to > the -march flag (--with-arch=armv7-a+fp). criu's Makefile explicitly > passes its own -march=armv7-a setting, which overrides the +fp > default, so we end up with no FPU: > cc1: error: '-mfloat-abi=hard': selected architecture lacks an FPU This was fixed with https://salsa.debian.org/debian/criu/-/commit/55d4dc414e6542c69c4fae95d31160256d726013 2/ https://github.com/openssl/openssl/issues/21630 (closed) The last comment is: > I've read the note several times but the behaviour of Configure does > not seem to match it. Rather than actually leaving the options open, > it adds the -march=armv7-a. If it had just picked the base linux-armv4 > target, things would have been fine. So, this is not going to be fixed in openssl. This could be worked around in ruby-build. Lucas
