If we allow __strub_leave to allocate a frame on sparc, it will overlap with a lot of the stack range we're supposed to scrub, because of the large fixed-size outgoing args and register save area. Unfortunately, setting up the PIC register seems to prevent the frame pointer from being omitted.
Since the strub runtime doesn't issue calls or use global variables, at least on sparc, disabling PIC to compile strub.c seems to do the right thing. Regstrapped on x86_64-linux-gnu, also testing on sparc-solaris2.11.3. (but it will likely take forever on the cfarm machine; if someone with faster sparc machines could give this a spin and confirm, that would be appreciated.) Ok to install? IIRC this patch gets 32-bit sparc to pass all strub tests, but sparc64 still fails many of them; there's another one for sparc64 that fixes them, and that will improve sparc -m32 as well. for libgcc/ChangeLog PR middle-end/112917 * config.host (sparc, sparc64): Enable... * config/sparc/t-sparc: ... this new fragment. --- libgcc/config.host | 2 ++ libgcc/config/sparc/t-sparc | 4 ++++ 2 files changed, 6 insertions(+) create mode 100644 libgcc/config/sparc/t-sparc diff --git a/libgcc/config.host b/libgcc/config.host index 694e3e9f54cad..54d06978a5d2c 100644 --- a/libgcc/config.host +++ b/libgcc/config.host @@ -199,9 +199,11 @@ riscv*-*-*) ;; sparc64*-*-*) cpu_type=sparc + tmake_file="${tmake_file} sparc/t-sparc" ;; sparc*-*-*) cpu_type=sparc + tmake_file="${tmake_file} sparc/t-sparc" ;; s390*-*-*) cpu_type=s390 diff --git a/libgcc/config/sparc/t-sparc b/libgcc/config/sparc/t-sparc new file mode 100644 index 0000000000000..fb1bf1fc29cc4 --- /dev/null +++ b/libgcc/config/sparc/t-sparc @@ -0,0 +1,4 @@ +# This is needed for __strub_leave to omit the frame pointer, without +# which it will allocate a register save area on the stack and leave +# it unscrubbed and most likely unused, because it's a leaf function. +CFLAGS-strub.c += -fno-PIC -fomit-frame-pointer -- Alexandre Oliva, happy hacker https://FSFLA.org/blogs/lxo/ Free Software Activist GNU Toolchain Engineer More tolerance and less prejudice are key for inclusion and diversity Excluding neuro-others for not behaving ""normal"" is *not* inclusive