[FFmpeg-devel] [PATCH] fate: use the lavfi -graph_file option to workaround path issues.

[Clément Bœsch]

On Wed, Oct 24, 2012 at 08:13:19PM -0300, jamal wrote:
> On 24/10/12 6:42 PM, Clément Bœsch wrote:
> > ---
> > FATE is still broken due to various path issues. This is an attempt to fix it
> > using the new introduced -graph_file option. I have no mean to test it with
> > mingw/msys/msvc stuff so I have no idea if it works as expected.
> > 
> > Here are the two issues supposed to be fixed:
> > 
> >   - the source file is passed as a standalone argument to printf, and I just
> >     hope here msys will do its work and replace paths such as "/d/foo/bar" into
> >     "d:/foo/bar". This is a problem currently happening with the scene
> >     detection because the filename is bogged down into the middle of the
> >     filtergraph and isn't substitute.
> 
> It's not doing it. scenedetect.lavfi contains the following for me when using SAMPLES=/ffmpeg/samples
> 
> sws_flags=+accurate_rnd+bitexact; movie='/ffmpeg/samples/svq3/Vertical400kbit.sorenson3.mov',select=gt(scene\,.4)
> 

Mmh, do you have any idea where/when the substitution is supposed to
happen? I thought MSYS was doing the replace from "/d/foo/bar" to
"d:\foo\bar" for each command argument, and since printf is just another
command…

Maybe it is using a shell built-in and thus would ignore the argument
search & replace? (is it better with a explicit /usr/bin/printf?)

Or I'm just completely mistaken?

> This aside, with and without this patch the only way i was able to make it accept the samples path and run without 
> failing was by giving it a relative instead of an absolute path (for example SAMPLES=../samples).

This is quite restrictive unfortunately, and I'm not sure FATE would
actually support relative paths that well.

> I remember having some issues compiling GCC when running the configure script using absolute paths in the past, so 
> i suppose it's a msys bug.
> 

I think I'm just doing it wrong here, but I must admit I have a hard time
trying to figure out how that really works since I don't have such
environment, and actually never used it.

> Personally, I'd say we should add a section to the fate documentation recommending msys users to use relative paths 
> for SAMPLES and not bloat the build system to work around a problem that should be fixed by the msys project.

It's not "that" bloating: the new introduced command actually simplifies
adding more metadata tests. Within that particular function we indeed have
to trick a little, but it doesn't even work well yet, so I believe we
should understand the problem a bit more before trying to even avoid it.

-- 
Clément B.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20121025/f235c2ed/attachment.asc>