On Tue, Apr 30, 2019 at 03:04:40PM +0200, Jeremie Courreges-Anglas wrote:
> On Tue, Apr 30 2019, Landry Breuil <landry@openbsd.org> wrote:
> > On Tue, Apr 30, 2019 at 01:49:49PM +0200, Landry Breuil wrote:
> >> On Sat, Apr 27, 2019 at 11:00:10AM +0200, Landry Breuil wrote:
> >> > On Wed, Apr 24, 2019 at 09:30:31AM -0600, Steven Mestdagh wrote:
> >> > > CVSROOT: /cvs
> >> > > Module name: ports
> >> > > Changes by: steven@cvs.openbsd.org 2019/04/24 09:30:31
> >> > >
> >> > > Modified files:
> >> > > math/lapack : Makefile distinfo
> >> > > math/lapack/files: Makefile
> >> > > math/lapack/pkg: PLIST
> >> > >
> >> > > Log message:
> >> > > update to 3.8.0
> >> >
> >> > It seems this has side-effects on py-numpy, which complains on a missing
> >> > symbol now:
> >> >
> >> > python2 -c 'import numpy'
> >> > python2:/usr/local/lib/liblapack.so.7.0: undefined symbol
> >> > 'ilaenv2stage_'
> >> >
> >> > same with python3, and this breaks geo/pdal configure via cmake (seen by
> >> > naddy@ and ajacoutot@)
> >> >
> >> > steven, can you investigate ?
> >>
> >> it also shows with rio from geo/rasterio:
> >>
> >> $rio
> >> python2.7:/usr/local/lib/liblapack.so.7.0: undefined symbol 'ilaenv2stage_'
> >>
> >> i had a quick look but my fortran fu is lacking.
> >>
> > sthen found https://bugzilla.redhat.com/show_bug.cgi?id=1514512 which
> > points at
> > https://src.fedoraproject.org/rpms/lapack/c/4ec46f8519f085bedb8ae9c16aa32b981e0d7004?branch=master
> > but we already have ilaenv2stage.o in ALLAUX in objdir/Makefile soooo i
> > dont see what can be wrong here. lld ?
>
> Our port uses its own Makefile instead of reusing upstream's build
> system so it needs to be kept in sync. Steven only missed one change in
> the update to 3.8.0:
Doh. How could i miss that... maybe we should just use upstream Makefile
then ?
> ok?
Definitely ok for me, thanks !
No comments:
Post a Comment