[FFmpeg-devel] WinCE/ffmpeg hacks

[İsmail Dönmez]

Hi,

On Wed, Jul 29, 2009 at 2:35 PM, Martin Storsj?<martin at martin.st> wrote:
> On Wed, 29 Jul 2009, Ismail D?nmez wrote:
>
>> On Wed, Jul 29, 2009 at 12:16 PM, Martin Storsj?<martin at martin.st> wrote:
>> > On Wed, 29 Jul 2009, Ismail D?nmez wrote:
>> >
>> >> On Wed, Jul 29, 2009 at 11:59 AM, Martin Storsj?<martin at martin.st> wrote:
>> >> >
>> >> > Do you have time to reply to my questions below, either privately or on
>> >> > the list? I'd like to know what differs in our setups, since you said that
>> >> > you only need to modify errno.h but nothing else than that. Are you able
>> >> > to build ffmpeg.exe and the mov demuxer with that setup?
>> >>
>> >> No, mov demuxer needs manual hacking since it uses sterror/errno
>> >> combo. And you are correct, you still need to hack ?strerror(),
>> >> perror(), getenv() and conio.h.
>> >
>> > Ok, good to know that our environments match in that respect, thanks.
>>
>> Looking at the hacks we need , we can't fix cegcc anymore because they
>> try to emulate WinCE and not the C standard. For the remaining
>> failures we need to add WinCE only hacks which defines these functions
>> to be NOP or such.
>
> Either that, or just keep local patches for the mingw32ce headers - I'm
> fine with that solution too, even though it isn't the cleanest possible
> one. With the attached patch (for mingw32ce 0.55), I'm able to build the
> latest ffmpeg unmodified.
>
> You could perhaps try arguing on their mailing list that supporting
> standard C headers would be wise even though they're not available in the
> official WinCE SDKs - that's quite easy as long as it only concerns
> headers and defines and functions that can be left with empty
> implementations, but for stuff like errno, doing it properly (with proper
> thread support) isn't really easy.
>
> Feel free to pursue this track (or trying to get workarounds merged into
> ffmpeg) if you like arguing, I think I'll settle for the local header
> hacks myself.

Thanks, atoll() one is fixed in cegcc SVN. For the rest I'll see what can I do.

Regards.

-- 
?smail D?NMEZ