Re: git: 27626082ac4c - main - emulators/dosbox-x: fix build on powerpc64*

Hello Piotr,

Thanks for commit.

Are these changes upstreamed?

Thanks,

Piotr Kubaj <pkubaj@freebsd.org> escreveu (sábado, 31/10/2026 à(s) 15:16):

> The branch main has been updated by pkubaj:
>
> URL:
> https://cgit.FreeBSD.org/ports/commit/?id=27626082ac4cc1c9d254fce0a4cd586c8ca27cca
>
> commit 27626082ac4cc1c9d254fce0a4cd586c8ca27cca
> Author:     Piotr Kubaj <pkubaj@FreeBSD.org>
> AuthorDate: 2026-10-27 18:49:49 +0000
> Commit:     Piotr Kubaj <pkubaj@FreeBSD.org>
> CommitDate: 2026-10-31 15:13:52 +0000
>
>     emulators/dosbox-x: fix build on powerpc64*
>
>     dosbox-x assumes long double is the x87 80-bit type on everything that
>     isn't x86 or ARM. It isn't on powerpc64*, so the FPU register union is
>     wrong there and 2026.10.01 doesn't even compile. Use the float80 path
>     like aarch64 does. Also fix the bundled lzma endianness check, FreeBSD
>     defines _LITTLE_ENDIAN on big-endian too.
> ---
>  emulators/dosbox-x/files/patch-include_dosbox.h              | 12
> ++++++++++++
>  .../dosbox-x/files/patch-src_libs_libchdr_lzma_CpuArch.h     | 10
> ++++++++++
>  2 files changed, 22 insertions(+)
>
> diff --git a/emulators/dosbox-x/files/patch-include_dosbox.h
> b/emulators/dosbox-x/files/patch-include_dosbox.h
> new file mode 100644
> index 000000000000..7685f059a3a0
> --- /dev/null
> +++ b/emulators/dosbox-x/files/patch-include_dosbox.h
> @@ -0,0 +1,12 @@
> +--- include/dosbox.h.orig      2026-10-27 11:58:28 UTC
> ++++ include/dosbox.h
> +@@ -104,6 +104,9 @@
> + #elif defined(__arm__) || defined(__aarch64__)
> + // ARMv7 (Raspberry Pi) does not have long double, sizeof(long double)
> == sizeof(double)
> + // ARM 64 has a quadruple-precision float instead of the 80-bit extended
> precision one used by x87
> ++#elif defined(__LDBL_MANT_DIG__) && __LDBL_MANT_DIG__ != 64
> ++// long double is not the x87 80-bit extended format: either the same as
> double (FreeBSD/powerpc64
> ++// and powerpc64le before 16.0) or IEEE binary128 (Linux ppc64le,
> riscv64, s390x, loongarch64)
> + #else
> + // GCC, other compilers, have sizeof(long double) == 10 80-bit IEEE
> + #define HAS_LONG_DOUBLE               1
> diff --git
> a/emulators/dosbox-x/files/patch-src_libs_libchdr_lzma_CpuArch.h
> b/emulators/dosbox-x/files/patch-src_libs_libchdr_lzma_CpuArch.h
> new file mode 100644
> index 000000000000..2f311f1d5d1b
> --- /dev/null
> +++ b/emulators/dosbox-x/files/patch-src_libs_libchdr_lzma_CpuArch.h
> @@ -0,0 +1,10 @@
> +--- src/libs/libchdr/lzma/CpuArch.h.orig       2026-01-14 10:48:42 UTC
> ++++ src/libs/libchdr/lzma/CpuArch.h
> +@@ -247,7 +247,6 @@ MY_CPU_64BIT means that processor can work with 64-bit
> +     || defined(MY_CPU_ARM_LE) \
> +     || defined(MY_CPU_ARM64_LE) \
> +     || defined(MY_CPU_IA64_LE) \
> +-    || defined(_LITTLE_ENDIAN) \
> +     || defined(__LITTLE_ENDIAN__) \
> +     || defined(__ARMEL__) \
> +     || defined(__THUMBEL__) \
>
>

-- 
Nuno Teixeira
FreeBSD UNIX:  <eduardo@FreeBSD.org>   Web:  https://FreeBSD.org