Bug 1712977

Summary:	test_gdb: test_gc() fails on s390x with: Unable to locate python frame
Product:	Red Hat Enterprise Linux 8	Reporter:	Charalampos Stratakis <cstratak>
Component:	python3	Assignee:	Victor Stinner <vstinner>
Status:	CLOSED ERRATA	QA Contact:	Lukáš Zachar <lzachar>
Severity:	unspecified	Docs Contact:
Priority:	low
Version:	8.1	CC:	cstratak, dmalcolm, extras-qa, mcyprian, mhroncok, pviktori, rkuska, shcherbina.iryna, tomspur, torsava, vstinner
Target Milestone:	rc
Target Release:	8.0
Hardware:	s390x
OS:	Unspecified
Whiteboard:
Fixed In Version:	python3-3.6.8-16.el8	Doc Type:	No Doc Update
Doc Text:		Story Points:	---
Clone Of:	1678277	Environment:
Last Closed:	2020-04-28 16:07:22 UTC	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:

Description Charalampos Stratakis 2019-05-22 15:31:25 UTC

+++ This bug was initially created as a clone of Bug #1678277 +++

The s390x job of the Fedora CI fails on building python3 package.

test_gdb.test_gc() failed twice on Python compiled in **debug mode** on s390x arch:

FAIL: test_gc (test.test_gdb.PyBtTests)
Verify that "py-bt" indicates if a thread is garbage-collecting
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/builddir/build/BUILD/Python-3.7.2/Lib/test/test_gdb.py", line 852, in test_gc
    self.assertIn('Garbage-collecting', gdb_output)
AssertionError: 'Garbage-collecting' not found in 'Breakpoint 1 (builtin_id) pending.\n[Thread debugging using libthread_db enabled]\nUsing host libthread_db library "/lib64/libthread_db.so.1".\n\nBreakpoint 1, builtin_id (self=, self@entry=<error reading variable: value has been optimized out>, v=42, v@entry=<error reading variable: value has been optimized out>) at /builddir/build/BUILD/Python-3.7.2/Python/bltinmodule.c:1216\n1216\t    return PyLong_FromVoidPtr(v);\nBreakpoint 2: file /builddir/build/BUILD/Python-3.7.2/Modules/gcmodule.c, line 860.\n[Inferior 1 (process 12917) exited normally]\nUnable to locate python frame\n'

Reformatted gdb catched output:

Breakpoint 1 (builtin_id) pending.
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".

Breakpoint 1, builtin_id (self=, self@entry=<error reading variable: value has been optimized out>, v=42, v@entry=<error reading variable: value has been optimized out>) at /builddir/build/BUILD/Python-3.7.2/Python/bltinmodule.c:1216
1216	    return PyLong_FromVoidPtr(v);
Breakpoint 2: file /builddir/build/BUILD/Python-3.7.2/Modules/gcmodule.c, line 860.
[Inferior 1 (process 12917) exited normally]
Unable to locate python frame

--- Additional comment from Victor Stinner on 2019-02-18 13:20:32 CET ---

Related bug reports:

* 2015: https://bugzilla.redhat.com/show_bug.cgi?id=1181034#c0 "test_gdb fails on s390(x)" closed as "EOL" (no fixed)
* 2016: https://bugzilla.redhat.com/show_bug.cgi?id=1333064#c0 "/CoreOS/python/Sanity/gdb test is failing on s390x" closed as ERRATA except that no change has been written :-)
* 2016: https://github.com/python/cpython/commit/0d88c652318c4c5eed77fd2e78e9e955ffa5c8d1 fix test_gdb on s390x (ignore some warnings)
* 2018: https://bugs.python.org/issue34007 "test_gdb fails in s390x SLES buildbots" fixed upstream in Python 3.7.1

--- Additional comment from Charalampos Stratakis on 2019-05-22 16:45:46 CEST ---

This seems to be the change that triggered the bug for some reason: https://src.fedoraproject.org/rpms/python3/c/8927d3f2671041d33c2682b47d238c7070474d0e?branch=master

Comment 1 Marcel Plch 2019-06-24 10:38:58 UTC

I have managed to reproduce the issue.
It seems like the compiler on s390x arch inlines functions even with -O0 (and therefore with -Og too) in CFLAGS, when certain flags are specified.

I'm trying to isolate the minimum set of flags required to reproduce the issue.

Comment 2 Marcel Plch 2019-07-19 11:51:48 UTC

The problem emerges from bad compiler flags handling.
I have submitted an upstream issue:
https://bugs.python.org/issue37631?

Comment 3 Charalampos Stratakis 2019-10-03 00:17:12 UTC

This issue can be fixed by making the following change to the SPEC:

-  make EXTRA_CFLAGS="$CFLAGS $MoreCFlags" %{?_smp_mflags}
+  %make_build CFLAGS_NODIST="$CFLAGS_NODIST $MoreCFlags"

This will make the -Og the be appended to the end of the compiler flags, since CFLAGS_NODIST are appended at the end as well. As it stands now, CFLAGS_NODIST contains the -O2 which overrides -Og for the debug build.

The same fix also resolves bug 1714733

Comment 4 Charalampos Stratakis 2019-10-03 01:10:14 UTC

As a side note, when trying to change -Og to -O0 the debug build, the compilation segfaults, as the hardening flags that are enabled by default, require optimizations higher than -O0

Comment 8 errata-xmlrpc 2020-04-28 16:07:22 UTC

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://access.redhat.com/errata/RHSA-2020:1764