https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124213
--- Comment #5 from Torbjorn SVENSSON <azoff at gcc dot gnu.org> --- After some more investigation, the bullet hole has been identified, but not yet the smoking gun. $ /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/arm-none-eabi-gcc -c -mfpu=auto -march=armv7-a+simd /tmp/t1.c /tmp/t2.c -### 2>&1 | grep -e '/as' -e '/cc1' /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/../libexec/gcc/arm-none-eabi/15.3.1/cc1 -quiet -imultilib thumb/v7-a/nofp -iprefix /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/../lib/gcc/arm-none-eabi/15.3.1/ -isysroot /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/../arm-none-eabi -D__USES_INITFINI__ /tmp/t1.c -quiet -dumpbase t1.c -dumpbase-ext .c "-mfpu=auto" "-mfloat-abi=soft" -marm "-mlibarch=armv7-a" "-march=armv7-a" -o /tmp/ccce7PAx.s /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/../lib/gcc/arm-none-eabi/15.3.1/../../../../arm-none-eabi/bin/as "-mfpu=softvfp" "-march=armv7-a" "-mfloat-abi=soft" "-meabi=5" -o t1.o /tmp/ccce7PAx.s /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/../libexec/gcc/arm-none-eabi/15.3.1/cc1 -quiet -imultilib thumb/v7-a/nofp -iprefix /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/../lib/gcc/arm-none-eabi/15.3.1/ -isysroot /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/../arm-none-eabi -D__USES_INITFINI__ /tmp/t2.c -quiet -dumpbase t2.c -dumpbase-ext .c "-mfloat-abi=soft" -marm "-mlibarch=armv7-a" "-march=armv7-a" -o /tmp/ccce7PAx.s /build/arm-gnu-toolchain-15.3.rel1-x86_64-arm-none-eabi/bin/../lib/gcc/arm-none-eabi/15.3.1/../../../../arm-none-eabi/bin/as "-mfpu=softvfp" "-march=armv7-a" "-mfloat-abi=soft" "-meabi=5" -o t2.o /tmp/ccce7PAx.s As can be seen, the -mfpu=auto option is only present for the first invocation of cc1.
