On 06/17/2016 05:10 PM, Richard Purdie wrote:
On Fri, 2016-06-17 at 14:42 -0700, Christopher Larson wrote:
On Fri, Jun 17, 2016 at 8:31 AM, Jolley, Stephen K <
stephen.k.jol...@intel.com> wrote:
· There is a multilib issue related to the layout of the
host libraries leaking into python’s build process which we’re
struggling to debug
Could I get some details on this? I've seen something like this where
cmake and automake are relying on sys.lib from python3-native/python
-native to determine the sitepackages directory, is that the behavior
others are hitting? So python modules end up in /usr/lib/ rather than
/usr/lib64/, for example?
If that's the issue in question, the issue is the mismatch between
the native sys.lib and target. We already patch automake to fall back
to using our libdir + python version to determine the path, but only
if it wasn't able to use sys.lib to do so. If we change that to
always use libdir+python version, it should fix the automake
packages. Then we might need to patch FindPythonLibs in cmake to do
the same. I'm testing the automake change locally now.
https://bugzilla.yoctoproject.org/show_bug.cgi?id=9717 is the bug
Proposed fixed for this problem:
http://lists.openembedded.org/pipermail/openembedded-core/2016-June/123061.html
Cheers,
Richard
--
_______________________________________________
Openembedded-core mailing list
Openembedded-core@lists.openembedded.org
http://lists.openembedded.org/mailman/listinfo/openembedded-core
