[SSSD] [PATCH] BUILD: Avoid symlinks with python modules

Jakub Hrozek jhrozek at redhat.com
Mon Oct 12 08:10:02 UTC 2015


On Mon, Oct 12, 2015 at 09:28:58AM +0200, Jakub Hrozek wrote:
> On Mon, Oct 12, 2015 at 09:26:12AM +0200, Jakub Hrozek wrote:
> > On Mon, Oct 12, 2015 at 07:54:45AM +0200, Lukas Slebodnik wrote:
> > > On (05/10/15 10:56), Lukas Slebodnik wrote:
> > > >ehlo,
> > > >
> > > >I had a discussion with debian maintainer on IRC:
> > > >04:07 < tjaalton> hmm, is adding symlinks to python modules a fedora thing? like pysss.so -> _py2sss.so
> > > >04:08 < tjaalton> looks like current debian python tools rename them, breaking the symlink
> > > >04:10 < lslebodn> it's not a fedora thing but it works on fedora
> > > >04:11 < lslebodn> It was the simplest way have to build python2 and python3 bindings in the same time
> > > >                  with automake
> > > >04:14 < tjaalton> ok
> > > >04:15 < lslebodn> _py2sssbecause there have to be the name of module in init function
> > > >04:15 < lslebodn> this is a reason why import _py2sss will fail
> > > >04:16 < lslebodn> if the debian python tool rename symbolic links then you can rename _py2sss.so after
> > > >                  installation
> > > >04:16 < lslebodn> mv _py2sss.so pysss.so
> > > >04:17 < lslebodn> instead of ln -s _py2sss.so pysss.so
> > > >04:18 < lslebodn> hmm, why I didn't do that in makefile? :-)
> > > >04:21 < lslebodn> tjaalton: If you have a probelm with symbolic links. You can try
> > > >                  http://paste.fedoraproject.org/273485/14436876
> > > >04:22 < lslebodn> but I didn't test the patch. I'm not sure if modules could be imported with such
> > > >                  change. But there shoudl not be a difference between mv and ln -s
> > > >04:24 < tjaalton> ok I'll try that, thanks
> > > >
> > > >Attached is an updated version of patch from previous fpaste link.
> > > >
> > > bump
> > > 
> > > LS
> > 
> > I didn't forget you, I just wasn't sure if pushing with failing Debian
> > tests is OK:
> >     http://sssd-ci.duckdns.org/logs/job/30/05/summary.html
> > 
> > But I guess since build passed it's fine..
> 
> Aaand now I remember that you said Timo confirmed this patch helped him
> which is a good substitute for Debian CI :-)
> 
> ACK

* master: bf0002fa68e9c06f0569cf75b1b8ef85bb1c2697
* sssd-1-13: 19181db613434db80e350d18c01221dcc1ff8a65


More information about the sssd-devel mailing list