[Swfdec] [PATCH] Fix logic in ffmpeg detection

[Benjamin Otte]

Due to overwhelming requests:

commit 35fdb57f33b63f078b0fa7062d0b970c7c739c7e
Author: Benjamin Otte <otte at gnome.org>
Date:   Sun Oct 28 22:49:46 2007 +0100

    disable ffmpeg and mad backends in default configure


Cheers,
Benjamin


On 10/25/07, Pavel Roskin <proski at gnu.org> wrote:
> Hello!
>
> On Thu, 2007-10-25 at 23:07 +0200, Benjamin Otte wrote:
> > Hi,
> >
> > On 10/25/07, Pavel Roskin <proski at gnu.org> wrote:
> > > If neither --enable-ffmpeg nor --disable-ffmpeg was given, try to find
> > > ffmpeg.  If it wasn't found, print the warning, but don't fail.
> > >
> > I've had had some discussions about automagic dependencies back in the
> > days with Mikel, and came to the conclusion that automagic is not a
> > good thing to do. A part of that (I think the non-relevant part) is in
> > bug 10333, I'll try to remember the important part:
> > The problem is that when you forget to install a dependency that you
> > would want to have, you will not notice (configure output is way too
> > noisy) unless configure aborts.
>
> Once possible solution would be to show the warning at the configure
> output and make it very visible.
>
> > This is why I decided to provide
> > default settings and enforce them unless they were overwritten via
> > configure. To be more exact, configure uses the suggested settings for
> > distributors while autogen.sh uses the suggested settings for
> > developers or people running through git.
>
> Distributors may be reluctant to use non-default options other than
> --prefix and similar stuff.  And that may be the reason why gnash is
> more popular with the distributions.  This is a shame, since swfdec
> seems to be more stable at the moment, at least for the casual end-user
> stuff (youtube, newgrounds.com etc).
>
> > So that's why I don't think that we should use automagic when
> > detecting audio backends. I'll also disable it for audio, when I
> > switch over to something more portable than alsa, like pulse. It's set
> > to automagic currently only to not annoy non-linux users.
> >
> > > If --enable-ffmpeg was given, fail if ffmpeg is not found.  If
> > > --disable-ffmpeg was given, don't print the warning, as the user already
> > > knows that ffmpeg won't be enabled.
> > >
> > That probably makes sense.
>
> Too bad it's the least important part of the patch :)
>
> > > It's important that ffmpeg is not viewed as a hard requirement, as it's
> > > absent in most distros due to patent problems.
> > >
> > Yeah, we should disable mad and ffmpeg by default. They provide
> > advantages to developers when debugging (like not spawning multiple
> > threads), but users shouldn't have to care. Currently I would not want
> > to do that, because mad and ffmpeg work better than GStreamer. This
> > might change in the near future though. Still, GStreamer would
> > definitely not be better technically, as it'd pretty much just be a
> > layer between mad and ffmpeg.
> >
> > A thing I'm wondering about is who views ffmpeg as a hard requirement
> > by looking at our configure.ac. And we've been very careful to mark
> > them as optional. The only hard dependencies of swfdec are cairo and
> > glib after all. Could you elaborate on that point?
>
> When I tried to compile swfdec, I saw the message, so I assumed that
> ffmpeg was required.  The standard behavior of the configure script is
> to abort only if the software cannot be compiled at all.
>
> I only discovered --without-ffmpeg today, much to my surprise.  After I
> sent the patch for ffmpeg, I removed all "-devel" packages that came
> from livna, and found that mad was another pseudo-requirement.
>
> My suggestion is that either the error message should suggest using the
> "--disable" option, or it should be a warning, perhaps at the end of the
> output.
>
> --
> Regards,
> Pavel Roskin
>
>