HOWTO: Install and use the latest FFmpeg and x264

[FakeOutdoorsman]

Note that i did aptitude hold, but that didnt stop apt-get apparently.. what does it stop? Aptitude?

I haven't used apt-get for a few years so I'm unsure if it has an equivalent of aptitude's "hold". I should have mentioned that hold probably would not stop apt-get and Synaptic.

With all that said, now my new compile is not working right. Im getting errors when using ffmpeg:

*** glibc detected *** ffmpeg: free(): invalid pointer: 0x08757158 ***
======= Backtrace: =========
/lib/tls/i686/cmov/libc.so.6[0xb7b5fa85]
/lib/tls/i686/cmov/libc.so.6(cfree+0x90)[0xb7b634f0]
ffmpeg(av_set_string2+0x8fb)[0x80eff9b]
======= Memory map: ========
08048000-08509000 r-xp 00000000 08:02 108607     /usr/local/bin/ffmpeg
08509000-08513000 rw-p 004c0000 08:02 108607     /usr/local/bin/ffmpeg

What is the ffmpeg command that you are using to get this error? I've never encountered this error, but I would like to try to reproduce it. My guess is that there may be a mess of several ffmpeg/x264 installs (or residual files) screwing each other up. Use you package manager to completely remove all installed yasm, x264, and ffmpeg. Now you can either:

cd ~/yasm
make distclean
./configure, make, checkinstall
cd ~/x264
make distclean
git pull
./configure, make, checkinstall
sudo ldconfig
cd ~/ffmpeg
make distclean
svn update
./configure, etc

or delete the yasm, ffmpeg and x264 directories and start the tutorial over.

[Leefmc]

or delete the yasm, ffmpeg and x264 directories and start the tutorial over.

I actually have, many times. I tried custom names, same error. Note that when i used custom names, upon removal, i made sure to sudo aptitude purge the names i gave them (and even the normal names, x264 && ffmpeg)

Last night i tried compiling it 4 to 6 different ways, each time i fully purged, and deleted, all of the installs.

I am out of ideas. Thoughts?

(Note i'll post the command i tried when i get home, but also note that the command worked great on my previous compile.)

[PGLGreg]

If you have an AMD 64-bit system, here is something I just stumbled on that might be useful. In compiling x264, the linker told me I should use "-fPIC" and recompile common/x86/cabac-a.o. But that .o file is assembled rather than compiled. So, hoping that the assembler (yasm) would understand, I started over and configured with:

./configure --prefix=/usr --libdir=/usr/lib64  \
  --enable-gtk --enable-pthread --enable-visualize --enable-pic --enable-shared  \
  --enable-mp4-output  \
  --extra-asflags="-fPIC"  \
  --extra-cflags="-O4 -march=k8 -mtune=k8 -pipe -ffast-math -fomit-frame-pointer"

That seems to have worked.

[FakeOutdoorsman]

I am out of ideas. Thoughts?

I suggest posting this problem to the ffmpeg-user mailing list. They are much more knowledgeable than I am with ffmpeg, but be sure to read the posting guidelines (first paragraph) or they may ignore or flame you if you don't follow their strict protocol. Also, this is unstated in their rules, but if you subscribe to the digest mode and reply to it with a non-mime digest compatible mail client then they will flame you. Make sure to include both your full ffmpeg command and the ffmpeg output or they will flame you.

[FakeOutdoorsman]

If you have an AMD 64-bit system, here is something I just stumbled on that might be useful. In compiling x264, the linker told me I should use "-fPIC" and recompile common/x86/cabac-a.o. But that .o file is assembled rather than compiled. So, hoping that the assembler (yasm) would understand, I started over and configured with:

./configure --prefix=/usr --libdir=/usr/lib64  \
  --enable-gtk --enable-pthread --enable-visualize --enable-pic --enable-shared  \
  --enable-mp4-output  \
  --extra-asflags="-fPIC"  \
  --extra-cflags="-O4 -march=k8 -mtune=k8 -pipe -ffast-math -fomit-frame-pointer"

That seems to have worked.

Thanks, PGLGreg. I'll add a note to the tutorial.

[Leefmc]

Ok, just got home. Seems this is my code:

ffmpeg -y -i $1 -pass 1 -threads 4 -s 1170x728 -vcodec libx264 -b 384k -bt 192k -f mp4 -flags +loop -cmp +chroma -partitions 0 -me_method epzs -subq 1 -trellis 0 -refs 1 -coder 0 -me_range 16 -g 300 -keyint_min 30 -sc_threshold 40 -i_qfactor 0.71 -maxrate 10M -bufsize 10M -rc_eq 'blurCplx^(1-qComp)' -qcomp 0.6 -qmin 10 -qmax 51 -qdiff 4 -level 30 -an -r ntsc /dev/null
ffmpeg -y -i $1 -pass 2 -threads 4 -s 1170x728 -vcodec libx264 -b 384k -bt 192k -f mp4 -flags +loop -cmp +chroma -partitions +parti4x4+partp4x4+partp8x8+partb8x8 -me_method umh -subq 7 -trellis 2 -refs 1 -coder 0 -me_range 16 -g 300 -keyint_min 30 -sc_threshold 40 -i_qfactor 0.71 -maxrate 10M -bufsize 10M -rc_eq 'blurCplx^(1-qComp)' -qcomp 0.6 -qmin 10 -qmax 51 -qdiff 4 -level 30 -acodec libfaac -ab 96k -ac 1 -r ntsc $2_temp.mp4

I forget where i got it from, but most of it is not by me. I tested a bunch, and eventually went with that one after tweaking it some to my liking. It produced good output, good filesize, and most of all it worked for QuickTime.

*edit*
Note that at the moment im taking random arguments from the commands and trying to make something work, to hopefully pinpoint the problem.

*edit2*
Best i can find, it seems this is the problem:

-rc_eq 'blurCplx^(1-qComp)'

If i remove that from the 2 commands, it runs. Im searching the ffmpeg docs to find out what rc_eq is.

*edit3*
Yea, seems it is a rate control equation. Not knowing encoding at all, how important is this? If i use the command "ffmpeg -rc_eq" i get the same error, so im guessing that specific command is corrupted or something? I compared an older videos size to the newly "no rc_eq" encoded one, and they seem to be the same size and quality, though i have not tested on quicktime yet (thats a bit of a pain when you dont have windows heh).

Thoughts?

[PGLGreg]

cd ~/x264
git pull
make distclean

Generally, you should clean up before doing the pull, because the pull might change the configuration in a way that confuses make and prevents it from doing the distclean correctly.

cd ~/x264
make distclean
git pull

[Nullack]

If you have an AMD 64-bit system, here is something I just stumbled on that might be useful. In compiling x264, the linker told me I should use "-fPIC" and recompile common/x86/cabac-a.o. But that .o file is assembled rather than compiled. So, hoping that the assembler (yasm) would understand, I started over and configured with:

./configure --prefix=/usr --libdir=/usr/lib64  \
  --enable-gtk --enable-pthread --enable-visualize --enable-pic --enable-shared  \
  --enable-mp4-output  \
  --extra-asflags="-fPIC"  \
  --extra-cflags="-O4 -march=k8 -mtune=k8 -pipe -ffast-math -fomit-frame-pointer"

That seems to have worked.

My understanding is that anything over -O3 in GCC just results in the compiler using 3.

I'm using -march=native on Intrepid, which is running well. I think native is a better setting.

[PGLGreg]

My understanding is that anything over -O3 in GCC just results in the compiler using 3.

I'm using -march=native on Intrepid, which is running well. I think native is a better setting.

Yes, that's my understanding, also. So, one may as well use "-O4".

I use those compiler flags not because I think they're good, but because mplayer thinks they're good. That is, when I configured mplayer, it took a look at my system and decided those are the flags I should use. Me -- I dunno.

What's good about "-march=native"? What does it even mean?

[FakeOutdoorsman]

*edit3*
Yea, seems it is a rate control equation.

I can confirm this error. I'm getting the same error but haven't found a reason or solution yet.