Re: [PATCH] [WIP] Port feature/native-comp to Windows.

[Andrea Corallo]

Andrea Corallo <address@hidden> writes:

> Nicolas Bértolo <address@hidden> writes:
>
>>> libdl exists only in MinGW64, I think.
>>
>>> Why do you need it? is it because the code uses dlopen/dlsym
>>> somewhere?
>>
>> The code uses `dynlib_open` and `dynlib_sym` which are implemented using 
>> libdl
>> in Posix systems. The original libgccjit uses it to implement the
>> `compile_to_memory` function that we do not use. The version of libgccjit I 
>> am
>> using implements that feature using native Windows functions, so libdl is not
>> necessary.
>>
>> I don't think we can remove it because that part of the code runs only when 
>> the
>> platform is not Mingw, and in that case libdl is necessary.
>>
>>> I think it's good to #undef each symbol before defining it.
>>
>> Fixed in the new patch.
>
> Hi Nicolas,
>
> I think would be handy to have 'native-comp-available-p' always defined
> also when compiling without the native compiler.  Similarly to what we
> have for gnutls-available-p and libxml-available-p.
>
> +static bool
> +load_gccjit_if_necessary(bool mandatory)
>                        ^^^
>                     space here
> +{
> +#ifdef WINDOWSNT
> +  if (!gccjit_initialized)
> +    {
> +      Lisp_Object status;
> +      gccjit_initialized = init_gccjit_functions ();
> +
>
> Also I think gccjit_initialized should be declared be static inside
> load_gccjit_if_necessary given should and is used only there.
>
> Thanks
>
>   Andrea

Also

+      gccjit_initialized = init_gccjit_functions ();
+      status = gccjit_initialized ? Qt : Qnil;
+      Vlibrary_cache = Fcons (Fcons (Qjson, status), Vlibrary_cache);

I think Qjson is a refuse from kill and yank.

Thanks

  Andrea

-- 
address@hidden