Re: [OpenBIOS] [PATCH] Fix cross compiling sparc64 on arm32 host

[Mark Cave-Ayland]

On 03/12/13 21:35, Cole Robinson wrote:
> Add arm32 to the whitelist for handling lack of int128 type variants.
> Otherwise we get errors like:
>
> /target/include/asm/types.h:79:1: error: unknown type name ‘__int128_t’
>      typedef __int128_t dcell;
> This may introduce issues building on aarch64 but I have no way to test.
>
> Recently, Fedora's build system had many arm hosts added. Since we always
> cross build openbios, the RPM build is scheduled independent of the
> build machine architecture. So whenever an openbios build was scheduled
> to an arm host, things fell over.
>
> config/scripts/switch-arch | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/config/scripts/switch-arch b/config/scripts/switch-arch
> index 44f2df4..86cab3c 100755
> --- a/config/scripts/switch-arch
> +++ b/config/scripts/switch-arch
> @@ -77,7 +77,7 @@ crosscflags()
>       if test "$target" = "sparc64" -o "$target" = "ia64" \
>            -o "$target" = "amd64" -o "$target" = "x86_64" \
>            -o "$target" = "alpha"; then
>    if test "$host" = "x86"; then
>    if test "$host" = "x86" -o "$host" = "arm"; then
>         cflags="$cflags -DNEED_FAKE_INT128_T"
>     elif test "$host" = "ppc" -a `uname -s` = "Darwin"; then
>         cflags="$cflags -DNEED_FAKE_INT128_T"

Hi Cole,

Thanks for the patch - I don't tend to do much with the build system but 
it looks fairly reasonable to me. One minor stylistic nit is that 
switch-arch seems to be grouped by arch, so would it be possible to 
rework so that the overall result looks like this:

if test "$host" = "x86"; then
if test "$host" = "arm"; then
     cflags="$cflags -DNEED_FAKE_INT128_T"
elif test "$host" = "ppc" -a `uname -s` = "Darwin"; then
     cflags="$cflags -DNEED_FAKE_INT128_T"

If you can resubmit like this (and also with a SOB) then I'm happy to 
apply it.

ATB,

Mark.

[PATCH] Fix cross compiling sparc64 on arm32 host

[Cole Robinson]

On 12/08/2013 06:57 AM, Mark Cave-Ayland wrote:
> On 03/12/13 21:35, Cole Robinson wrote:
> > Add arm32 to the whitelist for handling lack of int128 type variants.
> > Otherwise we get errors like:
> >
> > /target/include/asm/types.h:79:1: error: unknown type name ‘__int128_t’
> >      typedef __int128_t dcell;
> > This may introduce issues building on aarch64 but I have no way to test.
> >
> > Recently, Fedora's build system had many arm hosts added. Since we always
> > cross build openbios, the RPM build is scheduled independent of the
> > build machine architecture. So whenever an openbios build was scheduled
> > to an arm host, things fell over.
> >
> > config/scripts/switch-arch | 2 +-
> >   1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/config/scripts/switch-arch b/config/scripts/switch-arch
> > index 44f2df4..86cab3c 100755
> > --- a/config/scripts/switch-arch
> > +++ b/config/scripts/switch-arch
> > @@ -77,7 +77,7 @@ crosscflags()
> >       if test "$target" = "sparc64" -o "$target" = "ia64" \
> >            -o "$target" = "amd64" -o "$target" = "x86_64" \
> >            -o "$target" = "alpha"; then
> >    if test "$host" = "x86"; then
> >    if test "$host" = "x86" -o "$host" = "arm"; then
> >         cflags="$cflags -DNEED_FAKE_INT128_T"
> >     elif test "$host" = "ppc" -a `uname -s` = "Darwin"; then
> >         cflags="$cflags -DNEED_FAKE_INT128_T"
>
> Hi Cole,
>
> Thanks for the patch - I don't tend to do much with the build system but it
> looks fairly reasonable to me. One minor stylistic nit is that switch-arch
> seems to be grouped by arch, so would it be possible to rework so that the
> overall result looks like this:
>
> if test "$host" = "x86"; then
> if test "$host" = "arm"; then
>     cflags="$cflags -DNEED_FAKE_INT128_T"
> elif test "$host" = "ppc" -a `uname -s` = "Darwin"; then
>     cflags="$cflags -DNEED_FAKE_INT128_T"

Thanks for the review Mark, that above example code doesn't appear to work but
I've sent something similar.

- Cole

[PATCH] Fix cross compiling sparc64 on arm32 host

[Mark Cave-Ayland]

On 09/12/13 21:21, Cole Robinson wrote:
> Thanks for the review Mark, that above example code doesn't appear to work but
> I've sent something similar.
> Cole

Ooops yes, you've spotted the deliberate mistake. The new version looks 
fine to me, so I've applied it to SVN trunk.

Many thanks,

Mark.