Note: This is a public test instance of Red Hat Bugzilla. The data contained within is a snapshot of the live data so any changes you make will not be reflected in the production Bugzilla. Email is disabled so feel free to test any aspect of the site that you want. File any problems you find or give feedback at bugzilla.redhat.com.

Bug 1840077

Summary:	scipy test failures on s390x
Product:	[Fedora] Fedora	Reporter:	Miro Hrončok <mhroncok>
Component:	scipy	Assignee:	Nikola Forró <nforro>
Status:	ASSIGNED ---	QA Contact:	Fedora Extras Quality Assurance <extras-qa>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	33	CC:	cstratak, dan, python-sig, tomspur
Target Milestone:	---
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:		Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:	1853871
Bug Blocks:	467765

Description Miro Hrončok 2020-05-26 10:28:52 UTC

If you observe https://koschei.fedoraproject.org/package/scipy?collection=f33 you will see that ppc64le fails a lot.

With Python 3.9, I got ~600 test failures (~2% of collected tests). I've ended up pushing https://src.fedoraproject.org/rpms/scipy/c/3b286e1db7d0f46092207e6ac122ac19e7bd7a6a?branch=master as a dirty workaround. 

I've opened this to track the workaround, so we can eventually drop it. I have not analyzed the failures.

Side note: s390x tests are disabled entirely.

Comment 1 Ben Cotton 2020-08-11 13:34:18 UTC

This bug appears to have been reported against 'rawhide' during the Fedora 33 development cycle.
Changing version to 33.

Comment 2 Nikola Forró 2020-09-15 15:27:49 UTC

An update on this. With scipy 1.5.2, there are no failures on ppc64le, even with all the workarounds disabled. There is a single failure on armv7hl, which I'm currently investigating, and 11 failures on s390x (that's 0.03%, so I'm thinking about dropping the acceptable_failure_rate patch entirely).

Comment 3 Miro Hrončok 2020-09-15 17:59:29 UTC

Note that the latest build on rawhide n Koji (Fedora 34, scipy-1.5.0-4.fc34) has:

s390x: tests entirely skipped

armv7hl: 1 failed
i686: 2 failed

all others (incl. ppc64le): all passed


(Fedora 33 has similar results.)

So this is not actually blocked on 1.5.2 but it has somehow gone away naturally.

Comment 4 Nikola Forró 2020-09-30 19:31:09 UTC

I rebased scipy to 1.5.2, took care of the 2 tests failing (under right circumstainces) on 32-bit architectures, and explicitly disabled those 11 tests failing on s390x.
I'll try to fix those and keep this bugzilla open to track the progress.

Comment 5 Miro Hrončok 2020-09-30 21:04:49 UTC

Thanks!