Re: [PATCH] Make graphics/ffmpeg compile on loongson
On Fri, Oct 19, 2012 at 11:10:31PM +0200, Donovan Watteau wrote: On Sat, Sep 29, 2012 at 01:45:43PM +0200, Donovan Watteau wrote: Hello, The following diff makes it possible for ffmpeg to be linked on loongson (it used to give ld: final link failed: Bad value, see [1]), and probably sgi too (though I couldn't test this one). This is done by removing the call to -Wl,-Bsymbolic on mips targets. I'm not really sure to understand the effects of removing -Bsymbolic, but the resulting binary works fine for me on loongson. No REVISION bump since it only affects the archs on which it didn't build. [...] Could this go in, please? ffmpeg builds ok with this patch, thank you Donovan. Having ffmpeg on loongson is quite useful, IMHO, and it does unbreak mplayer, too (and possibly some other ports). I tried building vlc and mplayer from ports and neither build on my Fuloong. I need to test ffmpeg standalone since I can't get the above 2 working but I didn't want to hold this up since I can confirm ffmpeg builds cleanly after applying Donovan's patch and doesn't without it. My vlc build died after more than 30 hours(!) of attempting to build qt4 and deps. Don't try this at home, boys and girls. /jl -- ASCII ribbon campaign ( ) Powered by Lemote Fuloong against HTML e-mail X Loongson MIPS and OpenBSD and proprietary/ \http://www.mutt.org attachmentsCode Blue or Go Home!
Re: [PATCH] Make graphics/ffmpeg compile on loongson
On 10/27/2012 3:11 PM, John Long wrote: I tried building vlc and mplayer from ports and neither build on my Fuloong. I need to test ffmpeg standalone since I can't get the above 2 working but I didn't want to hold this up since I can confirm ffmpeg builds cleanly after applying Donovan's patch and doesn't without it. My vlc build died after more than 30 hours(!) of attempting to build qt4 and deps. Don't try this at home, boys and girls. /jl You will not get vlc working. QT4 doesn't work, even if you did get it to build. It actually will build successfully if you strip webkit-qt4 out but it doesn't work. I believe it's a binutils problem but I couldn't get anything useful to diagnose. ~Brian
Re: [PATCH] Make graphics/ffmpeg compile on loongson
On 10/27/2012 3:11 PM, John Long wrote: I tried building vlc and mplayer from ports and neither build on my Fuloong. I need to test ffmpeg standalone since I can't get the above 2 working but I didn't want to hold this up since I can confirm ffmpeg builds cleanly after applying Donovan's patch and doesn't without it. My vlc build died after more than 30 hours(!) of attempting to build qt4 and deps. Don't try this at home, boys and girls. /jl As an addendum, mplayer works just fine on my yeeloongs after building ffmpeg with Donovan's patch. ~Brian
Re: [PATCH] Make graphics/ffmpeg compile on loongson
On Sat, Sep 29, 2012 at 01:45:43PM +0200, Donovan Watteau wrote: Hello, The following diff makes it possible for ffmpeg to be linked on loongson (it used to give ld: final link failed: Bad value, see [1]), and probably sgi too (though I couldn't test this one). This is done by removing the call to -Wl,-Bsymbolic on mips targets. I'm not really sure to understand the effects of removing -Bsymbolic, but the resulting binary works fine for me on loongson. No REVISION bump since it only affects the archs on which it didn't build. [...] Could this go in, please? Having ffmpeg on loongson is quite useful, IMHO, and it does unbreak mplayer, too (and possibly some other ports). As for the proposed patch for GCC, it didn't work for me, see my previous message in this thread for more details. Thanks.
Re: [PATCH] Make graphics/ffmpeg compile on loongson
On Sat, 29 Sep 2012, Brad Smith wrote: This is a bug in the compiler. Looking around I believe the fix in this PR would resolve the issue. Try working with miod@ to have a bug fix applied for the compiler. http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33755 I recompiled GCC with the patch for GCC 4.2, and then I compiled ffmpeg again (without my patch for -Bsymbolic) but I still have the /usr/bin/ld: final link failed: Bad value error. I wonder whether I need to recompile ffmpeg's dependencies, though. I had tried a binutils-2.15 mipsel patch from Debian, some time ago while searching for this particular error, to no avail. But again, I was only recompiling ffmpeg after that.
Re: [PATCH] Make graphics/ffmpeg compile on loongson
On 9/29/2012 7:45 AM, Donovan Watteau wrote:
Hello,
The following diff makes it possible for ffmpeg to be linked
on loongson (it used to give ld: final link failed: Bad value,
see [1]), and probably sgi too (though I couldn't test this one).
This is done by removing the call to -Wl,-Bsymbolic on mips
targets. I'm not really sure to understand the effects of
removing -Bsymbolic, but the resulting binary works fine for me
on loongson.
No REVISION bump since it only affects the archs on which it
didn't build.
[1] http://marc.info/?l=openbsd-portsm=132691957332463w=2
Index: patches/patch-configure
===
RCS file: /cvs/ports/graphics/ffmpeg/patches/patch-configure,v
retrieving revision 1.31
diff -u -p -r1.31 patch-configure
--- patches/patch-configure 12 Sep 2012 08:15:57 - 1.31
+++ patches/patch-configure 29 Sep 2012 10:27:32 -
@@ -1,6 +1,6 @@
$OpenBSD: patch-configure,v 1.31 2012/09/12 08:15:57 ajacoutot Exp $
configure.orig Sun Jun 10 00:59:57 2012
-+++ configure Tue Sep 11 22:47:01 2012
+--- configure.orig Sun Jun 10 06:59:57 2012
configure Sat Sep 29 12:25:43 2012
@@ -1092,7 +1092,6 @@ HAVE_LIST=
lzo1x_999_compress
machine_ioctl_bt848_h
@@ -35,7 +35,17 @@ $OpenBSD: patch-configure,v 1.31 2012/09
enabled libmp3lame require libmp3lame = 3.98.3 lame/lame.h
lame_set_VBR_quality -lmp3lame
enabled libnut require libnut libnut.h nut_demuxer_init -lnut
enabled libopencore_amrnb require libopencore_amrnb
opencore-amrnb/interf_dec.h Decoder_Interface_init -lopencore-amrnb
-@@ -3075,10 +3075,6 @@ fi
+@@ -3058,7 +3058,8 @@ enabled extra_warnings check_cflags -Winline
+ # add some linker flags
+ check_ldflags -Wl,--warn-common
+ check_ldflags -Wl,-rpath-link,libpostproc -Wl,-rpath-link,libswscale
-Wl,-rpath-link,libavfilter -Wl,-rpath-link,libavdevice
-Wl,-rpath-link,libavformat -Wl,-rpath-link,libavcodec -Wl,-rpath-link,libavutil
+-test_ldflags -Wl,-Bsymbolic append SHFLAGS -Wl,-Bsymbolic
++# Avoid /usr/bin/ld: final link failed: Bad value with old binutils on mips
++! enabled mips test_ldflags -Wl,-Bsymbolic append SHFLAGS -Wl,-Bsymbolic
+
+ echo X{}; $TMPV
+ if test_ldflags -Wl,--version-script,$TMPV; then
+@@ -3075,10 +3076,6 @@ fi
if enabled small; then
add_cflags $size_cflags
@@ -46,7 +56,7 @@ $OpenBSD: patch-configure,v 1.31 2012/09
fi
check_cflags -fno-math-errno
check_cflags -fno-signed-zeros
-@@ -3506,6 +3502,7 @@ comment=$2
+@@ -3506,6 +3503,7 @@ comment=$2
version=$3
libs=$4
requires=$5
@@ -54,7 +64,7 @@ $OpenBSD: patch-configure,v 1.31 2012/09
enabled ${name#lib} || return 0
cat EOF $name/$name.pc
prefix=$prefix
-@@ -3519,7 +3516,7 @@ Version: $version
+@@ -3519,7 +3517,7 @@ Version: $version
Requires: $(enabled shared || echo $requires)
Requires.private: $(enabled shared echo $requires)
Conflicts:
@@ -63,7 +73,7 @@ $OpenBSD: patch-configure,v 1.31 2012/09
Libs.private: $(enabled shared echo $libs)
Cflags: -I\${includedir}
EOF
-@@ -3539,10 +3536,10 @@ Cflags: -I\${includedir}
+@@ -3539,10 +3537,10 @@ Cflags: -I\${includedir}
EOF
}
It's nice to have mplayer again without relying on the version packaged
with 5.0
~Brian
Re: [PATCH] Make graphics/ffmpeg compile on loongson
On Sat, Sep 29, 2012 at 01:45:43PM +0200, Donovan Watteau wrote: Hello, The following diff makes it possible for ffmpeg to be linked on loongson (it used to give ld: final link failed: Bad value, see [1]), and probably sgi too (though I couldn't test this one). This is done by removing the call to -Wl,-Bsymbolic on mips targets. I'm not really sure to understand the effects of removing -Bsymbolic, but the resulting binary works fine for me on loongson. No REVISION bump since it only affects the archs on which it didn't build. [1] http://marc.info/?l=openbsd-portsm=132691957332463w=2 This is a bug in the compiler. Looking around I believe the fix in this PR would resolve the issue. Try working with miod@ to have a bug fix applied for the compiler. http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33755 -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.
