Hi,

You aren't going to be able to avoid the glibc dependency. Most times
you can simply package (with cx_Freeze) the application on an "older"
machine that has all of the glibc requirements that you are looking
for. glibc is backwards compatible but not forwards compatible. That's
why I use CentOS 5.x for packaging instead of what I normally use
day-to-day (Fedora 13).

The only other option is to force the base executables to be built
statically and all Python extensions you are using to be included
statically as well -- judging from the list of extensions you are
using that isn't going to work for you. So see if you can get an "old"
distribution and build with that. I use a virtual machine (via Oracle
VirtualBox) for that purpose which works quite well.

Hope that helps you out.

Anthony

On Mon, Jul 12, 2010 at 12:25 PM, julian bilcke <julian.bil...@iscpif.fr> wrote:
> Hi !
>
> I have a new question regarding packaging of standalone Python application
> using cx_freeze.
>
> Just in preamble, a bit of context :
>
>  - We are building an open-source text-mining python server application for
> research, using a lot of existing python libs
>  - we have the following dependencies: Numpy, Twisted, (py-)BerkeleyDB,
> zope.interface, simplejson, yaml
>
> - Still, we have been asked to make it a "no additional bumper required"
> application (example of use cases: restrained environment without network
> connection, no admin access or install, or usb-key bootable app..)
>  - we still though this was feasible, so we decided to to build 4 versions
> (linux32, linux64, mac, win32)
> - We turned to tools like py2exe/app, pyinstall, and cx_freeze to make the
> job easier
>
> But this task appear to be difficult, because of these heavy dependencies
> (numpy, berkeley..)
>
> We have some results for Windows (binaries looks portable), but on other
> platforms, like mac or linux,
> this looks hard.
>
> By example, I tried to run our app under debian32 (it was built with
> cx_freeze under ubuntu32)
> but it failed : issue with numpy.so, which has an external dependency on
> glibc at a particular version.
> I don't have any idea on how resolve this kind of dependency without using
> the standard system packaging system..
>
> So, my question is :
>  - does someone had to do something similar ? is there a solution to this
> kind of issue (handling of external dependencies/libs) ?
> - it is even feasible to package such program ? or should we abandon and
> instead turn to a "local install" packaging ?
>
>
> if someone here had to deal with this kind of packaging, and has any hint, I
> would be very, very thankful !
>
> Julian
>
> ps:
>
> ------------------------------------------------------------------------------
> This SF.net email is sponsored by Sprint
> What will you do first with EVO, the first 4G phone?
> Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first
> _______________________________________________
> cx-freeze-users mailing list
> cx-freeze-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/cx-freeze-users
>
>

------------------------------------------------------------------------------
This SF.net email is sponsored by Sprint
What will you do first with EVO, the first 4G phone?
Visit sprint.com/first -- http://p.sf.net/sfu/sprint-com-first
_______________________________________________
cx-freeze-users mailing list
cx-freeze-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/cx-freeze-users

Reply via email to