Re: git: 01674e15dedb - main - bsd.endian.mk: Optimize the handling of big/little endian determination.

On 3 May 2026, at 03:24, Stephen J. Kiernan <stevek@freebsd.org> wrote:
> 
> The branch main has been updated by stevek:
> 
> URL: https://cgit.FreeBSD.org/src/commit/?id=01674e15dedb15f8ce498af0680c50bb075c5482
> 
> commit 01674e15dedb15f8ce498af0680c50bb075c5482
> Author:     Stephen J. Kiernan <stevek@FreeBSD.org>
> AuthorDate: 2024-04-03 23:25:19 +0000
> Commit:     Stephen J. Kiernan <stevek@FreeBSD.org>
> CommitDate: 2026-05-03 02:21:56 +0000
> 
>    bsd.endian.mk: Optimize the handling of big/little endian determination.
> 
>    Add variables to contain lists of MACHINE_ARCH values to use
>    to determine little or big endian.
> 
>    Only error out about not being able to determine endianess if
>    TARGET_ENDIANNESS is empty and not cross-compiling.
> 
>    Reviewed by:    sjg
>    Obtained from:  Hewlett Packard Enterprise Development LP
>    Differential Revision:  https://reviews.freebsd.org/D44629
> ---
> share/mk/bsd.endian.mk | 39 +++++++++++++++++++++++++++++----------
> 1 file changed, 29 insertions(+), 10 deletions(-)
> 
> diff --git a/share/mk/bsd.endian.mk b/share/mk/bsd.endian.mk
> index 0c5ebb7aeba2..2a617a1d2794 100644
> --- a/share/mk/bsd.endian.mk
> +++ b/share/mk/bsd.endian.mk
> @@ -1,20 +1,35 @@
> 
> -.if ${MACHINE_CPUARCH} == "aarch64" || \
> -    ${MACHINE_CPUARCH} == "arm" || \
> -    ${MACHINE_ARCH} == "amd64" || \
> -    ${MACHINE_ARCH} == "i386" || \
> -    ${MACHINE_ARCH} == "powerpc64le" || \
> -    ${MACHINE_CPUARCH} == "riscv"

Losing MACHINE_CPUARCH makes me sad as a downstream that adds new
MACHINE_ARCH for aarch64 and riscv. For riscv at least it’s now riscv*
which I guess is close enough, but for aarch64 it’s not wildcarded.
Similarly arm though I don’t care about that personally.

> +MACHINE_ARCH_LIST.little = \
> + aarch64 \
> + amd64 \
> + armv7 \
> + i386 \
> + powerpc64le \
> + riscv*
> +
> +MACHINE_ARCH_LIST.big = \
> + powerpc \
> + powerpc64
> +
> +.for e in big little
> +N_$e:= ${MACHINE_ARCH_LIST.$e:${M_ListToSkip}}

Should we be worried about catching MACHINE(_ARCH) being empty?

Jessica

> +.endfor
> +
> +# For the host, we need to look at the host architecture
> +.if ${MACHINE:Nhost*} == ""
> +_ENDIAN_ARCH=${_HOST_ARCH}
> +.else
> +_ENDIAN_ARCH=${MACHINE_ARCH}
> +.endif
> +
> +.if ${_ENDIAN_ARCH:${N_little}} == ""
> TARGET_ENDIANNESS= 1234
> CAP_MKDB_ENDIAN= -l
> LOCALEDEF_ENDIAN= -l
> -.elif ${MACHINE_ARCH} == "powerpc" || \
> -    ${MACHINE_ARCH} == "powerpc64"
> +.elif ${_ENDIAN_ARCH:${N_big}} == ""
> TARGET_ENDIANNESS= 4321
> CAP_MKDB_ENDIAN= -b
> LOCALEDEF_ENDIAN= -b
> -.elif ${.MAKE.OS} == "FreeBSD"
> -.error Don't know the endian of this architecture
> .else
> #
> # During bootstrapping on !FreeBSD OSes, we need to define some value.  Short of
> @@ -33,3 +48,7 @@ TARGET_ENDIANNESS= 1234
> CAP_MKDB_ENDIAN= -B # Poisoned value, invalid flags for both cap_mkdb
> LOCALEDEF_ENDIAN= -B # and localedef.
> .endif
> +
> +.if empty(TARGET_ENDIANNESS) && ${.MAKE.OS} == "FreeBSD"
> +.error Don't know the endianness of this architecture
> +.endif
>