Re: CVS commit: src

[Rin Okuyama <rokuyama.rk%gmail.com@localhost>]

Hi,

On 2025/09/11 14:16, Nathanial Sloss wrote:
> Module Name:	src
> Committed By:	nat
> Date:		Thu Sep 11 05:16:24 UTC 2025
>
> Modified Files:
> 	src/doc: HACKS
> 	src/tests/lib/libm: Makefile
>
> Log Message:
> Fix segfaults of libm tests noticable on m68k.
>
> Also document it in doc/HACKS.

...(snip)...
> +.if ${MACHINE} == "m68k" /* See doc/HACKS */
> +CPPFLAGS+= -O0
> +.endif
I don't understand this hack.

None of tests/lib/libm gets SEGV for me on my 68060 machine without
this hack. Here's outputs of atf-run(1) and atf-report(1):

https://www.netbsd.org/~rin/m68k-libm/atf-run_20250912.txt
https://www.netbsd.org/~rin/m68k-libm/atf-report_20250912.txt

Do you mean something like:

.if ${MACHINE} == "m68k" && ${MKSOFTFLOAT:Uno} != "no"

Even so, scope of hack seems too large for me. Usually, we apply
-O[01] hack for affected C sources:

COPTS.foo.c += -O0

If this does not work, there may be something more than "GCC bug".

Can you revert this hack at the moment, and reconsider please?

Thanks,
rin