On Wed, Nov 22 2017, "Kirill Bychkov" <[email protected]> wrote:
> On Wed, November 22, 2017 18:06, Jeremie Courreges-Anglas wrote:
>> On Wed, Nov 22 2017, "Kirill Bychkov" <[email protected]> wrote:
>>> On Tue, November 21, 2017 19:19, Jeremie Courreges-Anglas wrote:
>>>> On Tue, Nov 21 2017, Sol?ne Rapenne <[email protected]> wrote:
>>>>> Building games/tome4 on powerpc (macppc kernel) fail
>>>>>
>>>>> [-----]
>>>>> ==== Building luajit2 (release) ====
>>>>> Creating ../obj/Release/luajit2
>>>>> mkdir -p ../obj/Release/luajit2
>>>>> Running pre-build commands
>>>>> [---cut----]
>>>>> In file included from ../src/luajit2/src/lj_obj.h:14:0,
>>>>>                  from ../src/luajit2/src/lj_dispatch.c:9:
>>>>> ../src/luajit2/src/lj_arch.h:321:2: error: #error "No support for
>>>>> little-endian PowerPC"
>>>>>  #error "No support for little-endian PowerPC"
>>>>>   ^
>>>>>
>>>>> I propose to remove powerpc as a supported arch
>>>>
>>>> Fair proposal.  lang/luajit has a patch to explicitely disable this
>>>> error message, plus another one for the generated asm.
>>>>
>>>> What I don't understand, though, is why luajit thinks that powerpc is
>>>> little-endian: our powerpc architecture is definitely big-endian.
>>>>
>>>> Could you please try to run ''make test'' in lang/luajit on powerpc?
>>>
>>> Hi!
>>> I'm currently trying to build tome4 on macppc. It will take some more time
>>> on my single core G5 :)
>>> But... audio/openal uses COMPILER=base-clang ports-clang ports-gcc and have
>>> no chance to be built on powerpc because llvm is marked broken on that arch.
>>> It builds with ports-gcc.
>>
>> Yep, this is not nice.
>>
>> The problem that kills luajit and tome4 is silly: our sys/_endian.h
>> header always defines _LITTLE_ENDIAN and _BIG_ENDIAN.  To check
>> endianness, one should check the value of _BYTE_ORDER.
>>
>> (BYTE_ORDER if including <endian.h>)
>>
>> I don't know how upstream ended up with this additional check, but it
>> probably doesn't receive much attention, given that Linux defines
>> __LITTLE_ENDIAN, not _LITTLE_ENDIAN.  So who's actually building
>> little-endian powerpc programs?  I think it makes sense to just patch
>> out the offending test.
>
> I like the idea.

Looks like you already have a patch, then. :)

>> Then we can take a look at COMPILER ordering.
>> tome4 is probably not the only affected port...
>
> I'm sure there are plenty of them.

(See ports/infrastructure/mk/arch-defines.mk rev. 1.46, ports-clang
should now be skipped.)

> tome4 builds and starts (currently with a black screen) on my G5 with
> both patches taken from lang/luajit.

Well, if it's unusable, maybe it's not worth building it.

-- 
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF  DDCC 0DFA 74AE 1524 E7EE

Reply via email to