Re: [E-devel] EFL cross compilation for Windows

[Vincent T. <vin...@gm...>]

On Wed, Oct 26, 2016 at 10:34 AM, Andrii Kroitor <an....@sa...> wrote:
> Hello, Adrien.
>
> It's nice to meet you here.
>
> On 26.10.16 08:50, Adrien Nader wrote:
>> [...]
>>>> Should these packages be updated and uploaded manually?
>>> If the path with win builds does not work out we could host the windows
>>> binaries next to our source tarballs. Only for releases thought.
>> Writing and improving the package manager's GUI took time off the
>> packaging itself and has delayed the current release a lot.
> Could you share some estimation on release date, please? Have you changed
> or updated something in cross builds part?
> I have tried to setup win-builds cross compilation on few different
> Linux systems,
> but it was unsuccessful. Are your current scripts in public access
> somewhere?
>> But that actually doesn't matter much. It's supposed to be possible to
>> base on win-builds and build other versions or other packages and I
>> don't see why that couldn't apply to the EFLs.
>>
>> However I'm not sure I understand some of things done in that git
>> repository. I've only taken a cursory look at it and can't comment much
>> but I've been more than surprised to see a pthread.h file copied there
>> and that it came from pthreads-win32 which, I think(*), is
>> ABI-incompatible while winpthreasd doesn't have that issue (everyone and
>> their mothers assume pthread_t is an int like on linux).
>>
>> (*) maybe I've fixed the only occurrence of it a couple years ago, or
>> maybe not, I don't know anymore, but somehow I doubt it
> That file came from winpthreads-v3.3.0-2-x86_64-w64-mingw32.txz package
> from winbuilds.
> I've only added one explicit cast to bypass compilation error of
> upstream EFL.
> Problem is with pthread_cleanup_pop macro:
>
> #define pthread_cleanup_pop(E)\
>      (*pthread_getclean() = _pthread_cup.next,
> (E?_pthread_cup.func((pthread_once_t *)_pthread_cup.arg):0));}
>
> It works fine in C code, but in case of C++ last 0 should be casted to
> (void)

gustavo introduced that error, i've mentioned this bug to him. Instead
of adding a pthread implementation with potential problems with ABI,
fix really the problems in the EFL

Vincent