Bad interaction between signal and subprocess

[dvarrazzo]

BPO	2113
Nosy	@gpshead, @amauryfa, @dvarrazzo
Files	subprocess_signal_bug.py: Test case for the bug subprocess_signal_bug.patch: Proposed patch to fix the issue

^{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 = 'https://github.com/gpshead'
closed_at = <Date 2008-08-04.03:23:59.636>
created_at = <Date 2008-02-14.15:39:01.412>
labels = ['type-bug', 'library']
title = 'Bad interaction between signal and subprocess'
updated_at = <Date 2008-08-04.03:23:59.634>
user = 'https://github.com/dvarrazzo'

bugs.python.org fields:

activity = <Date 2008-08-04.03:23:59.634>
actor = 'gregory.p.smith'
assignee = 'gregory.p.smith'
closed = True
closed_date = <Date 2008-08-04.03:23:59.636>
closer = 'gregory.p.smith'
components = ['Library (Lib)']
creation = <Date 2008-02-14.15:39:01.412>
creator = 'piro'
dependencies = []
files = ['9431', '9432']
hgrepos = []
issue_num = 2113
keywords = ['patch']
message_count = 6.0
messages = ['62392', '62394', '62400', '64100', '69322', '70680']
nosy_count = 4.0
nosy_names = ['gregory.p.smith', 'jafo', 'amaury.forgeotdarc', 'piro']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue2113'
versions = ['Python 2.5']

[dvarrazzo]

During Popen.communicate(), if a signal is caught during the select(),
an unhandled exception is raised, and the output gathered is lost.

This means that a long running or hanged process can't be killed after a
timeout (as shown in the attached example, where the output collected
before the signal is valuable)

The bug happens only when stdout and stderr are not merged and is tested
on linux platform.

[dvarrazzo]

Just noticed that in the patch "except select.error:" suffices instead
of "except:"...

[amauryfa]

First, this patch has a possible bug: if select() fails on the first
iteration, wlist is left undefined, and the next instruction "if
self.stdin in wlist" will fail.

Second, I think that is is not a good idea to simply exit the loop on
the first signal. In your script, select() is interrupted because of
SIGALRM (try removing the call to os.kill), and there may be more data
to read.

A possible solution could be:

try:
...select.select()...
except select.error, e:
if e.args[0] == errno.EINTR:
continue # try again
else:
raise
else:
...exchange data...

[jafo]

Daniele: Please provide an updated version of this patch based on the
feedback below.

[gpshead]

Fixed as Amaury described in trunk r64756.

still needs backporting to release25-maint.

[gpshead]

fixed in release25-maint r65475.