Re: [oe-commits] Morgan Little : python: Update for python-native changes

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Tue, 2012-07-24 at 09:57 +0200, Martin Jansa wrote:
> On Mon, Jul 23, 2012 at 03:06:12PM +0200, Martin Jansa wrote:
> > On Mon, Jul 23, 2012 at 02:02:12PM +0100, Richard Purdie wrote:
> > > On Mon, 2012-07-23 at 13:37 +0200, Martin Jansa wrote:
> > > > On Mon, Jul 23, 2012 at 12:32:55PM +0100, Richard Purdie wrote:
> > > package-index doesn't inherit pythonnative. The python found from env
> > > would therefore be the host system python but its using files from the
> > > sysroot. Mixing the two is a bad idea.
> > > 
> > > The question is how is it mixing these up, the host system python
> > > shouldn't be looking there for modules.
> > > 
> > > So I'm not sure why they're getting mixed up but its likely the problem.
> > > Does anything jump out at you on the machine in question with the above
> > > in mind?
> > 
> > Not really and it happens only in some builds (all today - haven't seen
> > this error before) and sofar it failed in 4/10 image/feed builds and
> > worked in remaining 6.
> > 
> > Even for the same combination of MACHINE and target it once worked and once
> > failed without any metadata changes or host distro changes between.
> 
> env doesn't show anything interesting related to python
> 
> changing shebang to
> path/tmp-eglibc/sysroots/x86_64-linux/usr/bin/python-native/python
> 
> made all currently failing builds to succeed again, but the issue is
> still there..

Is this on the same build machine? There must be something in common
with the failed builds. Do they always fail if you
hardcode /usr/bin/python in there? If so, can we figure out why the host
python is looking at files in the sysroot? LD_LIBRARY_PATH isn't set or
something odd like that?

Cheers,

Richard