[MPlayer-dev-eng] SVN won't compile

[Kevin DeKorte]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 04/12/2009 04:26 PM, Diego Biurrun wrote:
> On Sun, Apr 12, 2009 at 03:23:55PM -0600, Kevin DeKorte wrote:
>> On 04/12/2009 12:14 PM, Diego Biurrun wrote:
>>> On Sun, Apr 12, 2009 at 11:36:09AM -0600, Kevin DeKorte wrote:
>>>> On 04/12/2009 09:31 AM, Diego Biurrun wrote:
>>>>> On Sun, Apr 12, 2009 at 08:29:32AM -0600, Kevin DeKorte wrote:
>>>>>> I've been getting this error the last couple of days when compiling svn
>>>>> I'm not getting this error.
>>>> Perhaps it is due to the fact that I'm compiling in 32bit mode on a
>>>> 64bit install.
>>>>
>>>> ./configure --target=i386-linux --cc="gcc -m32" --as="as --32"
>>>> - --prefix=/usr --
>>>> disable-nemesi --disable-vidix --disable-3dnow
>>>>
>>>> Worked fine on SVN r29156 just a couple of days ago.
>>> My sarcasm seems to have done the trick of making you provide some
>>> information, but it is still far from sufficient.  OS, CPU, compiler
>>> version are a strict minimum.  Also please try to find out which
>>> revision broke compilation for you.
>> Fedora 11 (rawhide)
>> CPU Q6600
>> gcc version 4.4.0 20090409 (Red Hat 4.4.0-0.32) (GCC)
>>
>> I switched to using an external a52 and the problem went away. Since
>> this appears to avoid that code altogether. The point here is, that if I
>> say --disable-3dnow, the 3dnow code should not be being pulled in, or at
>> least it should be ifdef'ed out. So that it doesn't compile.
> 
> The external liba52 lacks quite a few optimizations and should be
> slower.
> 
> Anyway, the problem must have been introduced by some of my recent work
> on configure.  I would like to know which commit exactly, so that I can
> fix it.  So please
> 
> - check if the problem persists with current svn and
> - try to find the exact revision that caused it.
> 
> Diego

Well I believe this is a gcc error, I went back several versions an
could not get the code to compile. Even to recreate a version I already
had. So I'll keep trying this when a new gcc comes out. It appears to be
an optimization error in gcc from what I can tell. But I could be way
off base.

I still think the code in imdct.c should respect the HAVE_AMD3DNOW that
is set in config.h, but it seems to want to enable it, no matter how you
have it set.

Kevin



- -- 
Get my public GnuPG key from
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0x7D0BD5D1
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/

iEYEARECAAYFAknjRZ4ACgkQ6w2kMH0L1dE49QCfY7U7GsEnYgehCCz21VgDToqk
rA4An3Sc4Gz5RWAVjL2lj0sKXEl1AkZQ
=ofdO
-----END PGP SIGNATURE-----