[MPlayer-dev-eng] mplayer-1.0rc2: fspp is broken

[Denys Vlasenko]

Hi Ivan,

Please CC me in replies, otherwise I might not notice
than a post is a reply to me. I am getting 400+ mails per day.

On Wednesday 30 April 2008 11:47, Ivan Kalvachev wrote:
> On Wed, Apr 30, 2008 at 11:14 AM, Denys Vlasenko
> <vda.linux at googlemail.com> wrote:
> > Hello,
> >
> >  mplayer-1.0rc1 was working ok for me. rc2 seems
> >  to have problems with fspp.
> >
> >  mplayer -vf fspp=5:1:6 movie.avi - produces loud
> >  hissing sound instead of normal audio (video seems
> >  to be ok).
> >
> >  mplayer -vf scale=1280 movie.avi - works as expected
> >  (movie is scaled up, sound is ok).
> >
> >  mplayer -vf scale=1280,fspp=5:1:6 movie.avi -
> >  again hissing sound, and the movie is not scaled!
> 
> First this is bugreport, and they don't belong to this maillist.

Done. Subscribed to mplayer-users now.

> Second, rc2 is 9 months old, there have been many fixes. Use current svn.

Downloaded 2008-05-02 snapshot. Build failed:

libass/ass_fontconfig.c: In function '_select_font':
libass/ass_fontconfig.c:148: error: 'FC_FULLNAME' undeclared (first use in this function)
libass/ass_fontconfig.c:148: error: (Each undeclared identifier is reported only once
libass/ass_fontconfig.c:148: error: for each function it appears in.)
libass/ass_fontconfig.c: In function 'validate_fname':
libass/ass_fontconfig.c:221: warning: passing argument 1 of 'utf8_get_char' from incompatible pointer type
make: *** [libass/ass_fontconfig.o] Error 1

It happens here:

        if (strcasecmp((const char*)val_s, family) != 0) {
                result = FcPatternGetString(rpat, FC_FULLNAME, 0, &val_s);
                if (result != FcResultMatch || strcasecmp((const char*)val_s, family) != 0)
                        mp_msg(MSGT_ASS, MSGL_WARN, MSGTR_LIBASS_SelectedFontFamilyIsNotTheRequestedOne,
                               (const char*)val_s, family);
        }

I grepped for FC_FULLNAME and it is never mentioned anywhere else.

I used configure options which are exactly equivalend to those which
I used for rc1 and rc2.

> Third, video causing audio issue indeed sounds like memory corruption
> (writing out of bounds), valgrind is rather good at detecting these. I
> find it strange gcc421 to cause stack align warning, as it is supposed
> to have fixed this issue.

My gcc did NOT emit stack align warning.
In my original post I said:

>  I googled for it and I am definitely not the only one
>  seeing this. Some people experience this and some are not.
>  For some, it seems compiler-dependent.
>
>  Unfortunately the most informative post I found still was
>  too terse:
>
>  "
>  Compiler did not align stack variables. Libavcodec has been miscompiled
>  and may be very slow or crash. This is not a bug in libavcodec,
>  but in the compiler. You may try recompiling using gcc >= 4.2.
>  Do not report crashes to FFmpeg developers.
>  "

In other words: someone else had a problem with fspp and _his_
gcc complained. Not mine. I don't know what version of gcc was that.
--
vda