> On 27 Dec 2021, at 05:24, Georgy Yakovlev wrote:
>
> otherwise initial build may fail with:
> unknown long double size, cannot define BFP_FMT
If possible, would you mind filing a bug with the build log
of it failing (and brief steps on how to get it) just so
we can easily test if we can drop it in future?
(We have a lot of hacks like this where I worry
we'll never be able to drop them with confidence)
If it takes a huge amount of work to get there, then don't
bother, as it's negligible gain.
>
> Signed-off-by: Georgy Yakovlev
> ---
> eclass/toolchain.eclass | 5 +
> 1 file changed, 5 insertions(+)
>
> diff --git a/eclass/toolchain.eclass b/eclass/toolchain.eclass
> index fd03ba176276..1102c4fc5d56 100644
> --- a/eclass/toolchain.eclass
> +++ b/eclass/toolchain.eclass
> @@ -1099,6 +1099,11 @@ toolchain_src_configure() {
> # Set up defaults based on current CFLAGS
> is-flagq -mfloat-gprs=double && confgcc+=( --enable-e500-double
> )
> [[ ${CTARGET//_/-} == *-e500v2-* ]] && confgcc+=(
> --enable-e500-double )
> + if [[ ${CTARGET} == powerpc-*-musl ]]; then
> + # fix: unknown long double size, cannot define BFP_FMT
> + confgcc+=( --disable-decimal-float )
> + export gcc_cv_target_ldbl128=no
> + fi
> ;;
> ppc64)
> # On ppc64 big endian target gcc assumes elfv1 by default,
> --
> 2.34.1
>
>
Looks fine otherwise though.
best,
sam
signature.asc
Description: Message signed with OpenPGP