latest hugin fc8 binary fails with segmentation fault
Tduell
I am using hugin-0.7.0-0.4.20080723svnfc8.x86_64 from Bruno's
repository.
I tried stitching a pair of scanned images, pretty much as per the
tutorial on that use of Hugin, but when it got to the final stitching
it failed with the following message...
nona -z PACKBITS -r ldr -m TIFF_m -o cath-pano -i 0 /tmp/
huginpto_4nrJ6y
nona -z PACKBITS -r ldr -m TIFF_m -o cath-pano -i 1 /tmp/
huginpto_4nrJ6y
enblend --compression NONE -f6114x2023 -o cath-pano.tif cath-
pano0000.tif cath-pano0001.tif
Loading next image: cath-pano0000.tif
Loading next image: cath-pano0001.tif
Creating blend mask: 1/4 2/4 3/4 4/4
Optimizing 1 distinct seam.
Strategy 1, s0: 1/4 2/4 3/4 4/4
Strategy 2: s0
make: *** [cath-pano.tif] Segmentation fault
make: *** Deleting file `cath-pano.tif'
The above is the complete text.
I tried this 3 more times. Once Hugin simply exited early in the
process without any message at all; one actually ran to completion but
produced an unsuitable result because I hadn't used the right
optimisation...I think that try was simply using the assistant to run
all the steps to see what would happen; and there was one other that
repeated the above message. I deleted the tep files between runs.
I think this is the first time I have this result, but it is also the
first time I have tried stitching scanned images.
Cheers,
Terry
Markku Kolkka
heinäkuuta 2008):
> Hullo All,
> I am using hugin-0.7.0-0.4.20080723svnfc8.x86_64 from Bruno's
> repository.
> I tried stitching a pair of scanned images, pretty much as per
> the tutorial on that use of Hugin, but when it got to the
> final stitching it failed with the following message...
>
> nona -z PACKBITS -r ldr -m TIFF_m -o cath-pano -i 0 /tmp/
> huginpto_4nrJ6y
> nona -z PACKBITS -r ldr -m TIFF_m -o cath-pano -i 1 /tmp/
> huginpto_4nrJ6y
> enblend --compression NONE -f6114x2023 -o cath-pano.tif cath-
> pano0000.tif cath-pano0001.tif
> Loading next image: cath-pano0000.tif
> Loading next image: cath-pano0001.tif
> Creating blend mask: 1/4 2/4 3/4 4/4
> Optimizing 1 distinct seam.
> Strategy 1, s0: 1/4 2/4 3/4 4/4
> Strategy 2: s0
> make: *** [cath-pano.tif] Segmentation fault
> make: *** Deleting file `cath-pano.tif'
>
I'm getting a similar segmentation fault in enblend with the
current fc8.x86_64 package (enblend-3.1-0.5.20080723cvs.fc8)
--
Markku Kolkka
markku...@iki.fi
Stefan F.
I built hugin, enblend etc. on 25 july from svn for ubuntu 8.04
greetings
Stefan
Bruno Postle
>
>> Creating blend mask: 1/4 2/4 3/4 4/4
>> Optimizing 1 distinct seam.
>> Strategy 1, s0: 1/4 2/4 3/4 4/4
>> Strategy 2: s0
>> make: *** [cath-pano.tif] Segmentation fault
>I'm getting a similar segmentation fault in enblend with the
>current fc8.x86_64 package (enblend-3.1-0.5.20080723cvs.fc8)
Yes current enblend CVS is broken (at least on x86_64 as the reports
I've seen are all on x86_64).
Regarding the fedora enblend package, I've removed the broken RPM
from the repository. Until I get a replacement set released, you
have to downgrade manually to the previous 20080714cvs RPM - The
easiest way to do this is to uninstall hugin/enblend and reinstall.
--
Bruno
Stefan F.
My CPU ist an x86 but I don't know whether it built a 64 or 32 bit
version.
Ubuntu's latest enblend package has version 3.0 and is therefore too
old.
How can I build the last stable enblend revision ?
Thanks
Stefan
Cyril Brulebois
> I was building enblend as described on
> http://wiki.panotools.org/Hugin_Compiling_ubuntu My CPU ist an x86 but
> I don't know whether it built a 64 or 32 bit version.
32-bit, unless it's actually x86_64, which would mean 64-bit.
> Ubuntu's latest enblend package has version 3.0 and is therefore too
> old.
>
> How can I build the last stable enblend revision ?
[1] might still help (we added a workaround for the --compression
feature, as documented in README.Debian, which you'll find in
/usr/share/doc/hugin, or online[1]).
1. http://packages.debian.org/sid/hugin
2. http://lists.alioth.debian.org/pipermail/pkg-phototools-devel/2008-July/000358.html
Feedback is very welcome.
Mraw,
KiBi.
Bruno Postle
Just follow the enblend compilation instructions except download
the last known 'good' CVS snapshot instead of the current 'head':
cvs -d:pserver:anon...@enblend.cvs.sourceforge.net:/cvsroot/enblend login
cvs -z3 -d:pserver:anon...@enblend.cvs.sourceforge.net:/cvsroot/enblend co -D '2008-07-18 21:00:00' -P enblend
--
Bruno
Stefan F.
> [1] might still help (we added a workaround for the --compression feature.
Do you mean I should install enblend 3.0 - only set compression to
NONE ?
----------
> the last known 'good' CVS snapshot instead of the current 'head':
I tried this, but get the same segmentation fault as in enblend of
july 25.
--------
> Yes current enblend CVS is broken (at least on x86_64 as the reports I've seen are all on x86_64).
type... i686-pc-linux-gnu) as I can't find any 64 bit switches in my
configuration.
The only 64 bit compiler switch is -D_FILE_OFFSET_BITS=64 which does
not seem to be a cpu specification.
cheers
Stefan
Cyril Brulebois
> On 27 Jul., 22:42, Cyril Brulebois wrote:
> > [1] might still help (we added a workaround for the --compression feature.
>
> Do you mean I should install enblend 3.0 - only set compression to
> NONE ?
I wasn't sure whether you were talking about the latest enblend release,
or the latest enblend CVS snapshot known to be working. In the former
case, I meant to point that enblend 3.0 is packaged, but I provided you
with an hugin link, sorry.
> I tried this, but get the same segmentation fault as in enblend of
> july 25.
>
> I guess enblend is also broken for normal 32bit systems (system
> type... i686-pc-linux-gnu) as I can't find any 64 bit switches in my
> configuration.
Maybe we could try a debug build (with -g, without stripping), and
provide a backtrace, so that people can compare them? Maybe they are
different issues after all.
> The only 64 bit compiler switch is -D_FILE_OFFSET_BITS=64 which does
> not seem to be a cpu specification.
Yep, that is LFS (Large File Support), needed to open large files.
Mraw,
KiBi.
Stefan F.
I forgott to install the new build - now everything ok!
For anybody trying to build from svn I recommend to prepare shell
scrips as below:
==========================
#! /bin/sh
# see http://wiki.panotools.org/Hugin_Compiling_ubuntu
cvs -d:pserver:anonym...@enblend.cvs.sourceforge.net:/cvsroot/enblend
login
enblend co -D '2008-07-18 21:00:00' -P enblend
echo "-------------------------------------------------------------"
read -p "Updated enblend - press any key." -r input
cd enblend
make -f Makefile.cvs
./configure
make
sudo make install
cd ..
echo ""
echo "-------------------------------------------------------------"
read -p "Built and installed enblend - press any key." -r input
#exit $errstatus
# build ends here
==========================
cheers
Stefan
Kornel Benko
> "Stefan F." <ste.f...@gmail.com> (28/07/2008):
> > On 27 Jul., 22:42, Cyril Brulebois wrote:
> > > [1] might still help (we added a workaround for the --compression feature.
> >
> > Do you mean I should install enblend 3.0 - only set compression to
> > NONE ?
>
> I wasn't sure whether you were talking about the latest enblend release,
> or the latest enblend CVS snapshot known to be working. In the former
> case, I meant to point that enblend 3.0 is packaged, but I provided you
> with an hugin link, sorry.
>
> > I tried this, but get the same segmentation fault as in enblend of
> > july 25.
> >
> > I guess enblend is also broken for normal 32bit systems (system
> > type... i686-pc-linux-gnu) as I can't find any 64 bit switches in my
> > configuration.
>
> Maybe we could try a debug build (with -g, without stripping), and
> provide a backtrace, so that people can compare them? Maybe they are
> different issues after all.
I tried. But then enblend did not crash.
Compile with "-g3" ==> no crash
Compile with "-g -O" ==> crash.
Debugging the second leads you deep into templates of "mask.h", but I was unable to see the cause.
One cannot step to the error case, because it takes ages, so I had to run it and wait for crash.
But then the stack does not look correct.
> > The only 64 bit compiler switch is -D_FILE_OFFSET_BITS=64 which does
> > not seem to be a cpu specification.
> Yep, that is LFS (Large File Support), needed to open large files.
>
> Mraw,
> KiBi.
>
Kornel
--
Kornel Benko
Kornel...@berlin.de
Tduell
On Jul 27, 7:34 am, Bruno Postle <br...@postle.net> wrote:
> Yes current enblend CVS is broken (at least on x86_64 as the reports
> I've seen are all on x86_64).
>
> Regarding the fedora enblend package, I've removed the broken RPM
> from the repository. Until I get a replacement set released, you
> have to downgrade manually to the previous 20080714cvs RPM - The
> easiest way to do this is to uninstall hugin/enblend and reinstall.
enblend-3.1-0.6.20080714cvs.fc8.x86_64.rpm
hugin-0.7.0-0.4.20080727svn.fc8.x86_64.rpm. Although the enblend is
still 20080714 it is now 3.1-0.6, so assume it is a 'fixed' build.
Anyway, I tried them with the same scanned image stitching project
that broke, and I can report that all worked as it should.
Cheers,
Terry