getting musepack-tools building on the mac again

[kuniklo]

I'm inclined to drop esd support on the mac for now. Most people won't have the esd libraries installed anyway and it's a pain installing all the fink stuff just to build it. Any opinions?

[Lefungus]

Esd is deprecated/obsolete anyway

[kuniklo]

Esd is deprecated/obsolete anyway

Good to know. Thanks.

I've actually decided to bite off a lot more and start writing autotools scripts for musepack-tools instead. I've got mppdec and replaygain building on the mac so far.

[xmixahlx]

great!

i scrapped my autotools build cuz i didn't think it would go anywhere... oops

from the 1.15s source we need mppenc/replaygain building

someone should look into converting mppdec to use libmusepack (or perhaps this should be the "break-out the frontend" stage, like below)

i've said this before in #mpc, but converting musepack to the flac/lame/unix development of library/frontend would be the most beneficial imho (other than sv8 obviously)

thanks for all the hard work K


mike

[kuniklo]

great!

i scrapped my autotools build cuz i didn't think it would go anywhere... oops

mike

This didn't turn out to be as hard as I thought it would be. I now have a working autotools build on both mac and linux that builds all three tools (mppenc/dec + replaygain).

You can grab the source via svn here:
http://www.caddr.com/svn/musepack-tools/trunk

Or as usual use websvn to grab it at:

http://www.caddr.com/websvn

The autotools scripts don't currently build static versions of the binaries but that can be fixed pretty easily.

[kuniklo]

I've compiled musepack-tools for Os X panther and packaged up the binaries here:

http://www.caddr.com/musepack-tools-1.15s-mac.tar.gz

If one of you admin types wants to grab this and put it up here somewhere I'd be much obliged.

[Shy]

Musepack.net is now updated with this bundle.

[ak]

You can grab the source via svn here:
http://www.caddr.com/svn/musepack-tools/trunk

Uhm, not quite mac related, but here's a little patch, if I may: http://4nykey.nm.ru/tmp/musepack-tools_min...ngw_cflags.diff

Basically this makes it to compile with mingw, plus some cflags clean-up:
-turn CFLAGS to AM_CFLAGS, so one can mess with 'em
'-finline-functions' removed, as it already included by -O3
'-funroll-all-loops' is kinda pointless, from what I've gathered
'-fstrength-reduce', '-frerun-loop-opt' and '-fexpensive-optimizations' are enabled at every optim level, except -O1

[kuniklo]

Thanks!

Are you sure about this part:

--- synthasm.nas (revision 36)
+++ synthasm.nas (working copy)
@@ -344,7 +344,7 @@
pushd ebx
mov ebx, [sp($buff0)]
mov ecx, [sp($V0)]
- mov edx, Di_opt
+ mov edx, Di_opt_SIMD
mov eax, 32
fld dword [bias3]

[ak]

Not quite, but otherwise there's unresolved symbol 'Di_opt', when linking mppdec.
Yeah and it's only needed for mingw.
I ran mppdec.exe through wine on couple of files, and output wav's are bit-identical to those outputted by linux compile, so at least it doesn't seem to break things.

[kuniklo]

Not quite, but otherwise there's unresolved symbol 'Di_opt', when linking mppdec.
Yeah and it's only needed for mingw.
I ran mppdec.exe through wine on couple of files, and output waves are bit-identical to those outputted by linux compile, so at least it doesn't seem to break things.

It looks like there's a Di_opt_SIMD declared in synthasm.nas and a Di_opt declared in synthtab.c and they both contain the same data, so I'm guessing the reference to plain Di_opt is linked to the synthtab.c definition, which doesn't work for some reason on mingw. I think it's harmless to switch it.

Thanks for the patch! I've applied it in svn.

[ak]

Thanks.

And while we're at it, here's two more
http://4nykey.nm.ru/tmp/musepack-tools_cfl...gsoverride.diff
adds cflags override for couple of files (as in original Makefile), frankly enough it causes a slight speed gain for both mppenc and mppdec.

http://4nykey.nm.ru/tmp/musepack-tools_pro...ofilenonag.diff
adds --enable-profile-info switch to ./configure (disabled by default), so mppenc will write Profile name instead of 'Unstable'.

[kuniklo]

Thanks for the patches. I'm on vacation right now but I'll apply these when I get back tomorrow.

Keep them coming!

[kuniklo]

Ok. I've applied both patches. Thanks again.

It looks like the build is now defaulting to a -O2 flag instead of the -O3 the original makefile used. Do you know a clean way to fix this without overwriting CFLAGS?

[ak]

Hi, thanks for commiting and sorry for delay.

Looks like AM_CFLAGS and that 'override' thing don't clap together very well.
I googled around, it appears forcing cflags is usualy done like this: http://4nykey.nm.ru/tmp/cflags_reorg.diff
Not sure how clean it is, wfm nevertheless.

[kuniklo]

Hi, thanks for commiting and sorry for delay.

Looks like AM_CFLAGS and that 'override' thing don't clap together very well.
I googled around, it appears forcing cflags is usualy done like this: http://4nykey.nm.ru/tmp/cflags_reorg.diff
Not sure how clean it is, wfm nevertheless.

Thanks. I'll take a look at this ASAP.