Bug 1606315

Created attachment 1467964 [details]
build.log

file build.log too big, will only attach last 32768 bytes

Created attachment 1467965 [details]
root.log

file root.log too big, will only attach last 32768 bytes

Created attachment 1467966 [details]
state.log

This bug appears to have been reported against 'rawhide' during the Fedora 29 development cycle.
Changing version to '29'.

Working on making this build on F30, a rebase to 1.2.0 seems to be sufficient.

It appears that scipy once again fails to build in rawhide.

In regards of python 3.8 rebuild, it is needed to recythonize the sources as proposed in https://src.fedoraproject.org/rpms/scipy/pull-request/7

Yet that makes the failure even worse.

BUILDSTDERR: Traceback (most recent call last):
BUILDSTDERR:   File "/usr/lib64/python3.7/site-packages/matplotlib/sphinxext/plot_directive.py", line 499, in run_code
BUILDSTDERR:     exec(code, ns)
BUILDSTDERR:   File "<string>", line 15, in <module>
BUILDSTDERR:   File "/builddir/build/BUILD/scipy-1.2.1/build/lib.linux-x86_64-3.7/scipy/interpolate/_bsplines.py", line 1017, in make_lsq_spline
BUILDSTDERR:     check_finite=check_finite)
BUILDSTDERR:   File "/builddir/build/BUILD/scipy-1.2.1/build/lib.linux-x86_64-3.7/scipy/linalg/decomp_cholesky.py", line 280, in cholesky_banded
BUILDSTDERR:     c, info = pbtrf(ab, lower=lower, overwrite_ab=overwrite_ab)
BUILDSTDERR: SystemError: <fortran object> returned NULL without setting an error
BUILDSTDERR: /usr/lib64/python3.7/site-packages/numpy/lib/utils.py:docstring of scipy.interpolate.spleval:15: WARNING: Block quote ends without a blank line; unexpected unindent.
BUILDSTDERR: /usr/lib64/python3.7/site-packages/numpy/lib/utils.py:docstring of scipy.interpolate.spline:11: WARNING: Block quote ends without a blank line; unexpected unindent.
BUILDSTDERR: /usr/lib64/python3.7/site-packages/numpy/lib/utils.py:docstring of scipy.interpolate.splmake:11: WARNING: Block quote ends without a blank line; unexpected unindent.
BUILDSTDERR: /builddir/build/BUILD/scipy-1.2.1/build/lib.linux-x86_64-3.7/scipy/misc/common.py:docstring of scipy.misc.ascent:28: WARNING: Exception occurred in plotting scipy-misc-ascent-1
BUILDSTDERR:  from /builddir/build/BUILD/scipy-1.2.1/doc/source/generated/scipy.misc.ascent.rst:
BUILDSTDERR: Traceback (most recent call last):
BUILDSTDERR:   File "/usr/lib64/python3.7/site-packages/matplotlib/sphinxext/plot_directive.py", line 499, in run_code
BUILDSTDERR:     exec(code, ns)
BUILDSTDERR:   File "<string>", line 2, in <module>
BUILDSTDERR:   File "/builddir/build/BUILD/scipy-1.2.1/build/lib.linux-x86_64-3.7/scipy/misc/common.py", line 157, in ascent
BUILDSTDERR:     with open(fname, 'rb') as f:
BUILDSTDERR: FileNotFoundError: [Errno 2] No such file or directory: '/builddir/build/BUILD/scipy-1.2.1/build/lib.linux-x86_64-3.7/scipy/misc/ascent.dat'

scipy blocks a large pile of packages, so this is a serious blocker for the 3.8 rebuilds.

Pull request created at: https://src.fedoraproject.org/rpms/scipy/pull-request/9

I have created a Dockerfile for OpenBLAS upstream to provide them with an environment where they can debug the issue with broken Fortran calls:
https://github.com/xianyi/OpenBLAS/issues/2137

Upstream claims that our build seems to cause the problem.

Setting needinfo on OpenBLAS Fedora maintainer.

Hello, Susi,
may I ask you to take a look at the upstream issue? Do you have any idea what can we do to resolve the issue?

Is this related to #1709161?
The broken linkage issue should be solved ASAP.

The Fedora package is standard OpenBLAS, so the problem can't originate from the packaging. (You may be using the wrong flavor though.)

It might be that this is the GCC bug discussed in
https://lwn.net/Articles/791393/

I've merged https://src.fedoraproject.org/rpms/scipy/pull-request/9