Closed GoogleCodeExporter closed 9 years ago
can you post system details and configure options used ?
thanks
Original comment by ya...@google.com
on 22 Feb 2014 at 7:19
OS: OpenBSD
Compiler / assembler: LLVM/clang
configure options utilized: --prefix=/usr/local --disable-optimizations
--disable-unit-tests --enable-shared
Original comment by brad.ope...@gmail.com
on 24 Feb 2014 at 9:34
Original comment by ya...@google.com
on 27 Feb 2014 at 11:08
I ran a clang build as follows:
CC=clang CXX=clang++ LD=clang++ ./configure --disable-optimizations
--disable-unit-tests --enable-shared
make clean && make -j12
clang --version
clang version 3.4 (trunk 187307)
Target: x86_64-unknown-linux-gnu
Thread model: posix
My OS: Ubuntu 4.4.3-4ubuntu5.1
This seemed to work properly.
Are you doing something different? Can you help me duplicate the problem?
Original comment by jimbankoski@google.com
on 28 Feb 2014 at 6:19
I am doing the same thing here and still see the same issue.
.e.g.
cd /tmp
$ tar -jxf /home/ports/distfiles/libvpx-v1.3.0.tar.bz2
$ cd libvpx-v1.3.0
CC=clang CXX=clang++ LD=clang++ ./configure --disable-optimizations
--disable-unit-tests --enable-shared
Configuring selected codecs
enabling vp8_encoder
enabling vp8_decoder
enabling vp9_encoder
enabling vp9_decoder
Configuring for target 'x86_64-linux-gcc'
enabling x86_64
enabling pic
enabling runtime_cpu_detect
enabling mmx
enabling sse
enabling sse2
enabling sse3
enabling ssse3
enabling sse4_1
enabling avx
enabling avx2
using yasm
checking here for x86inc "x86_64" "yes"
enabling use_x86inc
enabling postproc
Creating makefiles for x86_64-linux-gcc libs
Creating makefiles for x86_64-linux-gcc examples
Creating makefiles for x86_64-linux-gcc docs
gmake
*fails to build*
$ clang -v
clang version 3.5 (trunk)
Target: amd64-unknown-openbsd5.5
Thread model: posix
$ gmake -v
GNU Make 4.0
Built for x86_64-unknown-openbsd5.4
Copyright (C) 1988-2013 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Original comment by brad.ope...@gmail.com
on 1 Mar 2014 at 3:39
Also if you want access to this system to take a look around I can provide
access.
Original comment by brad.ope...@gmail.com
on 1 Mar 2014 at 3:43
Hmm when you do make can you do
make -p verbose=1 &> libvpx.log
and post the entire libvpx.log
and if you are cool with me ssh'ing into your machine I may give it a go. I
suggest a private email with your info.
Jim
Original comment by jimbankoski@google.com
on 2 Mar 2014 at 3:55
When I copy and paste what you have there but substitute make with gmake the
build never finishes and libvpx.log is empty. When the build gets to the point
where it "stops" I see on the screen... and I waited for awhile and then hit
Ctrl-C. Please e-mail me at brad at comstyle dot com and I'll provide the
relevant information to login to this system.
# No 'vpath' search paths.
# No general ('VPATH' variable) search path.
# strcache buffers: 4 (3) / strings = 3010 / storage = 28879 B / avg = 9 B
# current buf: size = 8162 B / used = 4414 B / count = 358 / avg = 12 B
# other used: total = 20717 B / count = 2652 / avg = 7 B
# other free: total = 21 B / max = 9 B / min = 5 B / avg = 7 B
# strcache performance: lookups = 3348 / hit rate = 10%
# hash-table stats:
# Load=3010/8192=37%, Rehash=0, Collisions=1164/3348=35%
# Finished Make data base on Sun Mar 2 18:17:14 2014
[1] + Done (2) gmake -p verbose=1
Original comment by brad.ope...@gmail.com
on 2 Mar 2014 at 11:27
Sorry I meant make -p verbose=1
Original comment by jimbankoski@google.com
on 2 Mar 2014 at 11:45
I think there is some confusion here. If I use "gmake -p verbose=1" and leave
off the rest of what you had specified in post #7 where does the logging come
into play? Since you did ask for the log file.
Original comment by brad.ope...@gmail.com
on 3 Mar 2014 at 12:16
Sorry maybe I should just try myself... ;)
Original comment by jimbankoski@google.com
on 3 Mar 2014 at 12:36
And as I said if you send me an e-mail to the address specified in post #8 I
will provide login information so you can play around on the system as you
please to be able to find out whatever other details that might be relevant
towards fixing the issue.
Original comment by brad.ope...@gmail.com
on 3 Mar 2014 at 12:38
Email sent
Original comment by jimbankoski@google.com
on 3 Mar 2014 at 12:43
Our vp8 assembly uses a trick so that we don't need to deal with structure
offsets in assemby by hard coding them.
To do so we generate inline assembly that sets up an offsets table using equ
statements:
We are using statements like this to do so:
asm("hi equ $4");
which is supposed to set a labelled address to value 4.
For some reason this is failing on your system and point to the dollar sign as
being wrong. It does not do this on mine.
I'm copying another person who might have more knowledge than I do, but at
this point I don't know what else to try.
Original comment by jimbankoski@google.com
on 3 Mar 2014 at 3:44
Any update on this?
Original comment by brad.ope...@gmail.com
on 7 Mar 2014 at 10:19
Ping. Between this and 724 it makes it impossible to move forward with libvpx
and although all the relevant apps will use FFmpeg for a VP9 decoder it would
be nice to be able to encode VP9 content too.
Original comment by brad.ope...@gmail.com
on 16 Mar 2014 at 11:58
Has this ever worked? Trying to decide if this is a regression, which would
allow for bisection.
If you are only interested in vp9, you can disable vp8. vp9 does not use the
offsets which appear to be giving you trouble.
Original comment by johannko...@google.com
on 16 Mar 2014 at 11:11
Looks like this doesn't work with v1.2.0. Anyway my use of Clang was to try and
workaround the issue pointed out with bug 724. If 724 could be fixed we would
be good to go.
No, I'm not just interested in VP9. We need the whole thing.
Original comment by brad.ope...@gmail.com
on 17 Mar 2014 at 12:48
Since 729 was closed and this is the same issue then this ticket might as well
be closed too.
Original comment by brad.ope...@gmail.com
on 11 Jul 2014 at 2:39
Original comment by johannko...@google.com
on 6 Nov 2014 at 1:23
Original issue reported on code.google.com by
brad.ope...@gmail.com
on 17 Feb 2014 at 9:26