[FFmpeg-devel] ARMv4 Development

[Asif Tasleem]

Hi Siarhei Siamashka,

The only thing I have changed in the code is the following function in
file.c:

static long lseekce (int fildes, long offset, int whence)
{
    DWORD mode = 0;
    switch (whence)
    {
      case SEEK_SET:
        mode = FILE_BEGIN;
        break;
      case SEEK_CUR:
        mode = FILE_CURRENT;
        break;
      case SEEK_END:
        mode = FILE_END;
        break;
    }
    return (long) SetFilePointer ((HANDLE) fildes, offset, NULL, mode);
}

As the lseek method is not supported on WinCE.

Could you please guide me how can I execute these test on ARMv6 device or
emulator?
I have not found enough help on Internet.
I have following environment available to me:

1- AT&T Tilt Device
2- MSYS/MinGW configured on Windows XP for Compilation
3- Microsoft Visual Studio 2005
4- Windows Mobile Device Emulator 3.0

Regards,
Asif Tasleem

On Wed, Jul 30, 2008 at 5:53 PM, Siarhei Siamashka <
siarhei.siamashka at gmail.com> wrote:

> On Monday 28 July 2008, Asif Tasleem wrote:
> > My team is working on a Windows Mobile 6 application that provides
> > multimedia processing functions like trimming and joining video files.
> >
> > I have compiled FFMPEG successfully for ARMv4 Processor.
> >
> > Following is the status of the operations that we are trying to perform
> on
> > windows Mobile.
> >
> > 1)      Trimming operation only trims the video data in a Video file.
> >
> > 2)   Changing -ab value to higher rates generates nothing in the output
> > file
> >
> > 3)      Discarding Audio/Video generates the files but files are not
> > playable.
> >
> > 4)      Conversion from AVI to MPEG is working 100%.
> >
> > 5)      Converting a sequence of images to a video crashes the
> application
> >
> >
> > However these operations are working 100% on the i386 compilation.
> >
> > Would you please like to refer some resource who could help us in finding
> > out the solution to these issues.
>
> If you have testcases where i386 build works fine while Windows Mobile 6
> build
> does not, you can try to look for the place where things start to be wrong.
> Enable the most verbose output, add some of the debugging output messages
> yourself. Once you find the offending part(s) of code, you can try to fix
> it
> and submit a patch or ask for further assistance.
>
> Standard FFmpeg regression test used to pass fine on ARMv6 not so long ago
> with ALT_BITSTREAM_READER. I may re-run it again shortly just to be sure.
>
> So while I'm basically only interested in audio/video decoding, encoding
> should not be in a very bad shape either. One of the possibilities
> is that you might have screwed something when building FFmpeg for Windows
> Mobile 6. Of course, bugs that are not caught by regression tests are also
> possible.
>
> As you are using an old ARM core, you may have some alignment problems not
> detected in my runs of regression tests on ARMv6 device. This is probably
> worth checking.
>
> > We are aimed to improve the FFMPEG for ARMv4 and definitely we will make
> > our efforts "open source" as well.
>
> Yes, making you "efforts" open source is the only legal way to use FFmpeg.
> That is if you add modifications to the code of course.
>
> --
> Best regards,
> Siarhei Siamashka
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at mplayerhq.hu
> https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel
>



-- 
Regards,
Asif Tasleem