Bugzilla Bug 234629

Arches,

Each profile needs entries added to package.use.mask for media-video/mplayer
plus the specific use flags, since the arch checks have been properly removed
from the ebuild.

All: win32codecs, real, custom-cpuopts

alpha, x86-fbsd, hppa, ia64, ppc64, sparc: cpudetection, dxr2, dxr3

Hrm? I don't see the arch checks of real and win32codecs removed.
Anyway, realplayer is a binary for x86 only, and win32codecs is for
alpha/amd64/ppc/x86 only, am i wrong?

As for custom-cpuopts and cpudetection, i'd say thats okay for all the arches.
About dxr2 and dxr3, those are for special graphics card, right? Doubt those
are available for other than amd64/ppc/x86 as well. At least the
alpha/ia64/sparc teams don't have the hardware to test.

(In reply to comment #1)
> Hrm? I don't see the arch checks of real and win32codecs removed.
> Anyway, realplayer is a binary for x86 only, and win32codecs is for
> alpha/amd64/ppc/x86 only, am i wrong?

They weren't in the use flags check, they were below in if statements.  Just
run a diff to previous version and you'll see the changes.  win32codecs and
32bit binary support is for x86/amd64 only

> As for custom-cpuopts and cpudetection, i'd say thats okay for all the arches.

custom-cpuopts only works for x86/amd64 CPU flag checks

You can test cpudetection and mask it if it doesn't work.

Error: Runtime CPU detection only works for x86, x86-64 and PPC!

So, as we have talked, win32codecs, real, custom-cpuopts and cpudetection
should be masked in the base profiles(base and embedded) and unmasked in those
arches that can run it.

For bsd:

cpudetection is fine
dxr3 already masked

dxr2: what's this ??? it needs dxr2ioctl.h which I've found there:
http://dxr2.sourceforge.net/projects/dxr2-driver/source/dxr2-driver-1.0.4.tar.gz
We used to have it, but see bug #153365
It fails on linux too. Please nuke that useflag.

custom-cpuopts: seems fine

win32codecs: was already unmasked here
but I see this in the ebuild:
    if ! use bindist && ! use real; then
        myconf="${myconf} $(use_enable win32codecs win32dll)"
    fi
wouldn't that disable win32codecs support at all (and thus real) for x86 users
if use=win32codecs real is specified ?

real: there is that in the ebuild:
    # Real codec support, only available on x86, amd64
    if use real && use x86; then
        myconf="${myconf} --realcodecsdir=/opt/RealPlayer/codecs"
    elif use real && use amd64; then
        myconf="${myconf} --realcodecsdir=/usr/$(get_libdir)/codecs"
    else
        myconf="${myconf} --disable-real"
    fi

which afaik wouldn't change anything if the real useflag is on or off unless we
are on x86 or amd64


for the real + win32codecs stuff also there is:
        x86? (
            win32codecs? ( media-libs/win32codecs )
            real? ( media-libs/win32codecs
                media-video/realplayer )
            )
        amd64? ( real? ( media-libs/amd64codecs ) )

in the deps, which doesn't really make sense if you want the useflag use.masked
instead of having arch dependant code.

As per bug #234775 we can add dirac and schroedinger useflags to the list for
some of you:
media-video/dirac
media-libs/schroedinger


Samples:
http://samples.mplayerhq.hu/V-codecs/Dirac/

This version does not seem to have dirac in ts support, so the samples from
http://dirac.kw.bbc.co.uk/download/video/notfinal/ would not work.


You can force the decoder:
mplayer -vc fflibdirac foo.avi
mplayer -vc fflibschroedinger foo.avi

Or you can encode some videos:
mencoder foo.avi -o bar.avi -oac mp3lame -ovc lavc -lavcopts
vcodec=vorbis:vcodec=libschroedinger

replace libschroedinger by libdirac for dirac.

libschroedinger is much faster but I tend to prefer the quality when encoded
with libdirac.

(In reply to comment #5)
> Or you can encode some videos:
> mencoder foo.avi -o bar.avi -oac mp3lame -ovc lavc -lavcopts
> vcodec=vorbis:vcodec=libschroedinger

hmm copy/paste error, this should be read as:
mencoder foo.avi -o bar.avi -oac mp3lame -ovc lavc -lavcopts
vcodec=libschroedinger

masked dxr2, dxr3, cpudetection, custom-cpuopts in ppc64 profiles and added
~ppc64 back to latest ebuild.

Marked ~hppa again. I don't agree with the mass masking that was needed, but
apparently even good ideas travel only so far.

(In reply to comment #3)
> Error: Runtime CPU detection only works for x86, x86-64 and PPC!
> 
> So, as we have talked, win32codecs, real, custom-cpuopts and cpudetection
> should be masked in the base profiles(base and embedded) and unmasked in those
> arches that can run it.
> 

Done.

And simpler, sorry for the confusion guys.