Repository navigation
buildbot: test_cmd_line failure on Tiger: [Errno 9] Bad file descriptor #52704
Description
| BPO | 8458 |
|---|---|
| Nosy | @gpshead, @db3l, @ronaldoussoren, @ncoghlan, @pitrou, @vstinner, @ned-deily |
Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.
Show more details
GitHub fields:
assignee = None
closed_at = <Date 2011-01-06.00:18:15.914>
created_at = <Date 2010-04-19.11:53:45.728>
labels = ['tests', 'type-bug', 'library']
title = 'buildbot: test_cmd_line failure on Tiger: [Errno 9] Bad file descriptor'
updated_at = <Date 2011-01-06.00:18:15.913>
user = 'https://github.com/vstinner'bugs.python.org fields:
activity = <Date 2011-01-06.00:18:15.913>
actor = 'pitrou'
assignee = 'none'
closed = True
closed_date = <Date 2011-01-06.00:18:15.914>
closer = 'pitrou'
components = ['Library (Lib)', 'Tests']
creation = <Date 2010-04-19.11:53:45.728>
creator = 'vstinner'
dependencies = []
files = []
hgrepos = []
issue_num = 8458
keywords = ['buildbot']
message_count = 14.0
messages = ['103583', '103584', '103598', '121275', '124320', '124321', '125260', '125263', '125277', '125306', '125352', '125354', '125357', '125486']
nosy_count = 7.0
nosy_names = ['gregory.p.smith', 'db3l', 'ronaldoussoren', 'ncoghlan', 'pitrou', 'vstinner', 'ned.deily']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue8458'
versions = ['Python 3.2']
Activity
vstinner commentedon Apr 19, 2010
The error only occured once, and it was not reproduced when test_cmd_line was rerunning in verbose mode.
http://www.python.org/dev/buildbot/builders/x86 Tiger 3.x/builds/25/steps/test/logs/stdio
Re-running test test_cmd_line in verbose mode
test_directories (test.test_cmd_line.CmdLineTest) ... ok
test_large_PYTHONPATH (test.test_cmd_line.CmdLineTest) ... ok
test_optimize (test.test_cmd_line.CmdLineTest) ... ok
test_q (test.test_cmd_line.CmdLineTest) ... ok
test_run_code (test.test_cmd_line.CmdLineTest) ... ok
test_run_module (test.test_cmd_line.CmdLineTest) ... ok
test_run_module_bug1764407 (test.test_cmd_line.CmdLineTest) ... ok
test_site_flag (test.test_cmd_line.CmdLineTest) ... ok
test_unbuffered_input (test.test_cmd_line.CmdLineTest) ... ok
test_unbuffered_output (test.test_cmd_line.CmdLineTest) ... ok
test_usage (test.test_cmd_line.CmdLineTest) ... ok
test_verbose (test.test_cmd_line.CmdLineTest) ... ok
test_version (test.test_cmd_line.CmdLineTest) ... ok
----------------------------------------------------------------------
Ran 13 tests in 3.860s
OK
vstinner commentedon Apr 19, 2010
See also bpo-8457.
ned-deily commentedon Apr 19, 2010
I have also seen this test failure occasionally but only on OS X 10.4 systems, not 10.5 or 10.6.
ned-deily commentedon Nov 16, 2010
With 3.2a4, OS X 10.4 Tiger buildbot failures reported:
======================================================================
ERROR: test_run_code (test.test_cmd_line.CmdLineTest)
----------------------------------------------------------------------
======================================================================
ERROR: test_run_module (test.test_cmd_line.CmdLineTest)
----------------------------------------------------------------------
======================================================================
ERROR: test_version (test.test_cmd_line.CmdLineTest)
----------------------------------------------------------------------
On my 10.4 system, it seems to be a race condition: sometimes I see all three of these failures, sometimes just one, sometimes none. Again, only on 10.4 (Tiger), not 10.5 or 10.6. But the 10.4 machine I'm using is by far the slowest of the three so it is possible that could be a factor. Perhaps there is a race condition with cleaning up the p2c pipe from a previous run?
ned-deily commentedon Dec 18, 2010
See also msg124254.
pitrou commentedon Dec 18, 2010
This would be rather strange, since it would have a different file descriptor. Or it means there's a bug somewgere (perhaps in subprocess, or perhaps even in io.open).
It someone can reproduce even intermittently on their home machine, it would be nice to dig a bit deeper.
gpshead commentedon Jan 4, 2011
The code surrounding the p2c file descriptors in subprocess has changed a bit recently since this issue was filed. someone still needs to come up with a way to reproduce it.
pitrou commentedon Jan 4, 2011
Well, it occurred today again :)
http://www.python.org/dev/buildbot/all/builders/x86%20Tiger%203.x/builds/1884/steps/test/logs/stdio
But I would suggest an OS X bug instead. My random guess is that the child process exits very quickly (as you see, it runs "python -c" which is an immediate error), therefore the other end of the pipe gets closed and OS X then returns EBADF when we try to do something with our end of the pipe.
gpshead commentedon Jan 4, 2011
that race condition, if thats what this is, is likely not OS X specific. the child process could complete or close its fds before we reach that code on any OS. We either need to guard these io.open(p2c*...) lines against EBADF OSError's or we should create the io wrappers before _execute_child() is called.
pitrou commentedon Jan 4, 2011
Creating them before _execute_child() is the obvious and simple solution
(if my diagnosis is correct). Is there any downside to doing so?
gpshead commentedon Jan 4, 2011
Off the top of my head I don't _think_ so.. at least for posix. I'm not sure about the implications of the if mswindows: code in there though.
pitrou commentedon Jan 4, 2011
Well we could try a run of it on the buildbots, then :)
pitrou commentedon Jan 4, 2011
Trying it in r87744.
pitrou commentedon Jan 6, 2011
I'm closing it in the hope that it's fixed. If it reappears, we'll have to reopen.