On Sun, Aug 04 2019, Jeremie Courreges-Anglas <j...@openbsd.org> wrote:
> On Wed, Jul 31 2019, Charlene Wendling <juliana...@posteo.jp> wrote:
>> We're currently trying to build godot on these 2 archs for nothing and
>> it's pretty time consuming: ~6 hours on a PowerBook 5,8.
>>
>>> http://build-failures.rhaalovely.net/powerpc/2019-07-14/games/godot.log
>>
>> There were occurrences of NULL that gcc does not consider as boolean
>> false as you can read in the log,
>
> This would be an error on all architectures where ports-gcc is used, not
> just powerpc or sparc64.
>
>> that was easy to fix, but then:
>>
>> core/safe_refcount.h:126: undefined reference to `__sync_add_and_fetch_8'
>>
>> (...and more). We don't support the 64-bit __sync_* family, only the
>> newer __atomic_* one. So i've marked it as BROKEN like the others.
>>
>>> http://build-failures.rhaalovely.net/sparc64/2019-07-11/games/godot.log
>>
>> The log says that the architecture is not supported.
>
> This one looks a bit dumb. The #ifdef "arch" block doesn't define
> a macro common to all known archs so the #error doesn't make sense.
>
>> OK?
>
> I'd rather have something like NOT_FOR_ARCHS = ${GCC49_ARCHS} if we
> don't fix the NULL vs false mismatch.
>
> I'm running a build with the diff below, just to see if there are other
> problems.
There are no other problems, godot packages fine on sparc64 with that
diff. Do you want to try to add -latomic to LINKFLAGS on powerpc/hppa,
or shall we mark godot as BROKEN there, as you suggested?
> But maybe it would be better to restrict godot to
> architectures where it can actually be used...
>
>
> Index: patches/patch-servers_visual_shader_language_cpp
> ===================================================================
> RCS file: patches/patch-servers_visual_shader_language_cpp
> diff -N patches/patch-servers_visual_shader_language_cpp
> --- /dev/null 1 Jan 1970 00:00:00 -0000
> +++ patches/patch-servers_visual_shader_language_cpp 4 Aug 2019 13:47:47
> -0000
> @@ -0,0 +1,21 @@
> +$OpenBSD$
> +
> +error: converting to 'bool' from 'std::nullptr_t' requires
> direct-initialization
> +
> +Index: servers/visual/shader_language.cpp
> +--- servers/visual/shader_language.cpp.orig
> ++++ servers/visual/shader_language.cpp
> +@@ -1869,11 +1869,11 @@ const ShaderLanguage::BuiltinFuncDef
> ShaderLanguage::b
> +
> + bool ShaderLanguage::_validate_function_call(BlockNode *p_block,
> OperatorNode *p_func, DataType *r_ret_type) {
> +
> +- ERR_FAIL_COND_V(p_func->op != OP_CALL && p_func->op != OP_CONSTRUCT,
> NULL);
> ++ ERR_FAIL_COND_V(p_func->op != OP_CALL && p_func->op != OP_CONSTRUCT,
> false);
> +
> + Vector<DataType> args;
> +
> +- ERR_FAIL_COND_V(p_func->arguments[0]->type != Node::TYPE_VARIABLE,
> NULL);
> ++ ERR_FAIL_COND_V(p_func->arguments[0]->type != Node::TYPE_VARIABLE,
> false);
> +
> + StringName name = static_cast<VariableNode
> *>(p_func->arguments[0])->name.operator String();
> +
> Index: patches/patch-thirdparty_thekla_atlas_nvcore_nvcore_h
> ===================================================================
> RCS file: patches/patch-thirdparty_thekla_atlas_nvcore_nvcore_h
> diff -N patches/patch-thirdparty_thekla_atlas_nvcore_nvcore_h
> --- /dev/null 1 Jan 1970 00:00:00 -0000
> +++ patches/patch-thirdparty_thekla_atlas_nvcore_nvcore_h 4 Aug 2019
> 13:47:47 -0000
> @@ -0,0 +1,14 @@
> +$OpenBSD$
> +
> +Index: thirdparty/thekla_atlas/nvcore/nvcore.h
> +--- thirdparty/thekla_atlas/nvcore/nvcore.h.orig
> ++++ thirdparty/thekla_atlas/nvcore/nvcore.h
> +@@ -115,7 +115,7 @@
> + #elif defined POSH_CPU_STRONGARM
> + # define NV_CPU_ARM 1
> + #else
> +-# error "Unsupported CPU"
> ++//# error "Unsupported CPU"
> + #endif
> +
> +
--
jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE
