[MPlayer-dev-eng] [PATCH] resubmit realv4.diff

[Balatoni Denes]

Hi!


> 0x303?5652 should perhaps be replaced with #defined values or add some
> rv_version variable perhaps, yes its allready there before your patch but
> it makes reviewing the patch more difficult

I will change that to the fourcc macro Arpi proposed. Probably in this 
century, when I have time.

> the kf change from *2 to *4 looks like some additional bit has been
> inserted, in which case this would be wrong
could be. well the patch is an enhancement to rv30/40 demuxing, but of course
still not perfect (and I am not sure it will ever be). i thought if it's in 
the repository others can look at it too, and improve it. Some (most??) 
rv30/40 videos have correct a-v sync with the patch. rv20/10 is worse than 
before in some cases (choppy, but I actually didn't find it that bad, ymmv)

Sure there is huge room to improve still.


> +  if( ( ( s[0] & (~0xf) ) ==0 ) || ( ( s[0] & (~0xf) ) ==0x40 ) ||
> +     ( ( s[0] & (~0xf) ) ==0x50 ) || ( ( s[0] & (~0x7) ) ==0x10 ) ){
> this should be replaced by a simple 2 bit check as arpi said
okay. like
(s[0]<0x18) || ((s[0]>>5)==2)   ?

but keep in mind, that these could change, I only looked at the samples on the 
ftp, and of course I have no idea what bits do what. it works for me.

---btw if you care to recall, these were the values I found for s[0] :
1x0? i/p
<0x18 i/p
>=0x18 b
didn't see any 0x2? or 0x3?
and for rv40 in addition:
0x4? i/p      
0x50 i/p
0x6? b
<0x78 b
----

> why not
>                        int fps2= stream_read_word(demuxer->stream);
>                        if (fps2>0) sh->fps= fps2;
>                        sh->frametime = 1.0f/sh->fps;
looks better.

all in all imho this patch does more good than harm
and as a matter of fact with the newer streams this demuxer will surely break.

bye
Denes