[FFmpeg-devel] [PATCH 15/17] fate: update refs for AVCPBProperties fields change after bump

[Anton Khirnov]

Quoting Andreas Rheinhardt (2021-04-05 17:34:42)
> Anton Khirnov:
> > ---
> >  tests/ref/fate/mxf-d10-user-comments | 2 +-
> >  tests/ref/fate/ts-demux              | 2 +-
> >  2 files changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/tests/ref/fate/mxf-d10-user-comments b/tests/ref/fate/mxf-d10-user-comments
> > index 13761fb0ce..609271ac05 100644
> > --- a/tests/ref/fate/mxf-d10-user-comments
> > +++ b/tests/ref/fate/mxf-d10-user-comments
> > @@ -6,7 +6,7 @@
> >  #codec_id 0: mpeg2video
> >  #dimensions 0: 1280x720
> >  #sar 0: 3/4
> > -0,         -1,          0,        1,   150000, 0x0547870d, S=1,       24, 0x5aa90ad0
> > +0,         -1,          0,        1,   150000, 0x0547870d, S=1,       40, 0x7ea50ad0
> >  0,          0,          1,        1,   150000, 0xe80a1612, F=0x0
> >  0,          1,          2,        1,   150000, 0xc5c50e2f, F=0x0
> >  0,          2,          3,        1,   150000, 0x51e28a04, F=0x0
> > diff --git a/tests/ref/fate/ts-demux b/tests/ref/fate/ts-demux
> > index cdf34d6af0..c20364483b 100644
> > --- a/tests/ref/fate/ts-demux
> > +++ b/tests/ref/fate/ts-demux
> > @@ -15,7 +15,7 @@
> >  1,       5760,       5760,     2880,     1536, 0xbab5129c
> >  1,       8640,       8640,     2880,     1536, 0x602f034b, S=1,        1, 0x00bd00bd
> >  1,      11520,      11520,     2880,      906, 0x69cdcbcd
> > -0,      32037,      36541,     1501,   114336, 0x37a215a8, S=2,        1, 0x00e000e0,       24, 0x663d0b52
> > +0,      32037,      36541,     1501,   114336, 0x37a215a8, S=2,        1, 0x00e000e0,       40, 0x91e10b52
> >  0,      33538,      33538,     1501,    12560, 0xb559a3d4, F=0x0, S=1,        1, 0x00e000e0
> >  0,      35040,      35040,     1501,    12704, 0x2614adf4, F=0x0, S=1,        1, 0x00e000e0
> >  0,      36541,      41046,     1501,    51976, 0x9ff1dbfe, F=0x0, S=1,        1, 0x00e000e0
> > 
> There is a problem here: The alignment of int64_t, uint64_t is only four
> for 32bit x86 (and probably other arches); as a consequence,
> sizeof(AVCPBProperties) == 36 (not 40) and the checksums are different.
> One way to fix this would be to switch the size field to 64bits; a more
> hacky way would be to add a dummy element.

I said this before - IMO we should just stop doing side data checksums
here. They make no sense - side data is simply not guaranteed to have
the same in-memory representation everywhere.

Bending public API to work around broken tests is really not the way to go.

-- 
Anton Khirnov