[issue30263] regrtest: log the system load?

STINNER Victor report at bugs.python.org
Wed May 3 19:09:31 EDT 2017


New submission from STINNER Victor:

It's more and more common to get test failures only on a small group of buildbots, or even a single buildbot. In many cases, it's a race condition which depends on the system load. The system load depends on how many tests are running in parallel, if the buildbot slave allows multiple builds in parallel, etc. Sometimes, the failing test pass when it's run sequentially at the end of regrtest, sometimes it fails again.

I propose to add the system load after the timestamp to ease debug. Example of output on a patched regrtest when tests are run sequentially:

0:19:37 load: 1.28 [304/405/1] test_sunau
0:19:37 load: 1.28 [305/405/1] test_ctypes
...
0:20:02 load: 1.86 [309/405/1] test__osx_support -- test_socketserver skipped (resource denied)
...
0:20:09 load: 1.79 [314/405/1] test_int_literal
0:20:09 load: 1.79 [315/405/1] test_queue
0:20:13 load: 1.73 [316/405/1] test_fork1
0:20:20 load: 1.67 [317/405/1] test_atexit
...
0:22:09 load: 1.16 [346/405/1] test_sys_setprofile
0:22:09 load: 1.16 [347/405/1] test_selectors
0:22:24 load: 0.98 [348/405/1] test_bufio
0:22:25 load: 0.98 [349/405/1] test_string_literals

----------
components: Tests
messages: 292944
nosy: haypo
priority: normal
severity: normal
status: open
title: regrtest: log the system load?
type: enhancement
versions: Python 3.7

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue30263>
_______________________________________


More information about the Python-bugs-list mailing list