← Back to team overview

pyexiv2-developers team mailing list archive

Re: Introduction to my aims... and query on a glitch...

 

Rob,

I don’t think it has anything to do with boost’s version, although I may
be wrong. If Philip’s feedback helps us understand what exactly the
cause of the problem is, that would be great.

On 2010-09-05, Rob Healey <robhealey1@xxxxxxxxx> wrote:
> Dear Phillip:
> 
> Are you using linux or windows?  If Linux, what distribution are you
> using please?  I am using Fedora 15/ rawhide using boost-1.44.0...

>From the logs attached, Philip is running Ubuntu, Karmic I would guess
from the versions of the packages.

> I have no problems, but using boost-1.41, I had the same problems as you....
> 
> Sincerely yours,
> Rob G. Healey

Cheers,

Olivier


> On Sun, Sep 5, 2010 at 4:44 AM, Olivier Tilloy <olivier@xxxxxxxxxx
> <mailto:olivier@xxxxxxxxxx>> wrote:
> 
>     Hi Philip,
> 
>     On 2010-09-03, Philip Graham <philip.graham567@xxxxxxxxx
>     <mailto:philip.graham567@xxxxxxxxx>> wrote:
>     > Hi guys,
>     > The first time around after install and getting around the initial
>     "mt"
>     > error, running python in the terminal, I got the error "undefined
>     > symbol: _ZNK5Exiv27ExifKey3keyEv" when trying to import the pyexiv2
>     > module the very first time.
> 
>     I have never observed this issue myself, but I know that Rob Healey has,
>     and it seems that upgrading dependencies fixed the issue somehow.
> 
>     > This error is unintelligible to me as I am not familiar with c++
>     > programming. I am attracted to the Python bindings as I am
>     attempting to
>     > build an app for myself with pygtk to replace "f-spot" (see note
>     below:).
>     > Retracing my steps of preparation, I found that I had unintentionally
>     > built and installed exiv2-0.20 being the most recent on that site. I
>     > have now uninstalled exiv2-0.20 and installed a fresh build of
>     exiv2-0.19.
>     > After again running the scons and then the scons install against the
>     > exiv2-0.19 version, then running python in the terminal, I still
>     get the
>     > error "undefined symbol: _ZN5Exiv25ErrorD1Ev" (slightly different)
>     when
>     > trying to import the pyexiv2 module.
> 
>     From the list of packages installed on your system, you have exiv2 0.19
>     (0.19-1), but libexiv2 0.18 (0.18.2-1). That’s suspicious. It may very
>     well be that pyexiv2 was compiled against one version of libexiv2, and
>     another version is loaded when trying to import pyexiv2, causing the
>     error.
> 
>     I would try to completely uninstall all version of libexiv2 (including
>     those compiled and installed locally, e.g. in /usr/local/), install
>     either 0.19 or 0.20 (either should work just fine), and retry compiling
>     pyexiv2. Note that the latest version is available in
>     pyexiv2-developers’ PPA at
>     https://launchpad.net/~pyexiv2-developers/+archive/ppa
>     <https://launchpad.net/%7Epyexiv2-developers/+archive/ppa> (for
>     Lucid and
>     Maverick).
> 
>     > Are there some rogue files left over from the previous install?
>     How do I
>     > uninstall all of pyexiv2 any version and then re-install? Is it just a
>     > matter of deleting the pyexiv directory and the shared object file
>     > libexiv2python.so in /usr/lib/python2.6/dist-packages/...?
> 
>     Yes, that should be enough to completely uninstall pyexiv2 from your
>     system. The build log you attached lists all the files being installed,
>     and there’s no additional magic involved. Note that you can install
>     pyexiv2 locally (for the current user only, and without the need for
>     admin privileges): `scons install --user`.
> 
>     > I have included my scons build log, the error text when in Python,
>     > details of the packages installed on my system, summary of my system.
> 
>     I hope this helps, please let us know if it does solve the issue, and
>     don’t hesitate to ask here for any question related to building/using
>     pyexiv2.
> 
>     > Background note:
>     > I am digitising a legacy collection of 10,000 Kodachrome_25 35mm
>     > transparencies with original dates that are meaningless when displayed
>     > inside a modern app like "f-spot" or "digicam". I wish to have the
>     > option of displaying the thumbnails and previews in filename order as
>     > this is how they have been named and saved when manually scanned
>     in high
>     > definition raw mode. The date of creation when scanned is
>     meaningless in
>     > the timelines as used in apps receiving downloads from digital
>     cameras.
>     > But I was discouraged when I looked at the C# code in "f-spot" as to
>     > what to change to create the filename thumbnail display to govern an
>     > order of display. In the scans I embed full descriptions of subject
>     > matter along with the copyright and date definitions using the
>     power of
>     > EXIF tagging. Presently I use the perl-based 'ExifTool' on the command
>     > line, but I would really like to do this under Python hence my
>     interest
>     > in in your excellent pyexiv2.
> 
>     That should be pretty straightforward to achieve with pyexiv2.
>     As an anecdote, I started this project because I wanted to fix the
>     DateTime tags in pictures that either didn’t have one or that had a
>     incorrect one because the photographer couldn’t be bothered setting the
>     date/time on his camera correctly…
> 
>     > I do not understand the obsession with timelines in the current
>     crop of
>     > digital photo apps. Surely there are enough of the large number of
>     > colour transparencies being scanned in to warrant the option of
>     > filename-based displays in suitable display-management apps?
> 
>     Those apps are as good as what their authors want them to do… And it
>     seems that in this field every single person has a different opinion on
>     the best way of organizing their pictures collection. I myself haven’t
>     found the perfect application to suit my needs…
> 
>     > Greetings from the newbie
>     >
>     > -PGG-
> 
>     Cheers,
> 
>     Olivier



References