multiprocessing: conn_recv_string() broken error handling

[hfuru]

BPO	10133
Nosy	@vstinner
Files	conn_recv_string_failures.diff: conn_recv_string() error handling

^{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 2012-06-11.15:12:45.652>
created_at = <Date 2010-10-17.20:14:54.053>
labels = ['type-bug', 'expert-IO']
title = 'multiprocessing: conn_recv_string() broken error handling'
updated_at = <Date 2012-06-12.10:35:26.590>
user = 'https://bugs.python.org/hfuru'

bugs.python.org fields:

activity = <Date 2012-06-12.10:35:26.590>
actor = 'hfuru'
assignee = 'jnoller'
closed = True
closed_date = <Date 2012-06-11.15:12:45.652>
closer = 'sbt'
components = ['IO']
creation = <Date 2010-10-17.20:14:54.053>
creator = 'hfuru'
dependencies = []
files = ['19256']
hgrepos = []
issue_num = 10133
keywords = ['patch']
message_count = 4.0
messages = ['118978', '162618', '162622', '162665']
nosy_count = 6.0
nosy_names = ['hfuru', 'vstinner', 'jnoller', 'asksol', 'python-dev', 'sbt']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue10133'
versions = ['Python 3.1', 'Python 2.7', 'Python 3.2']

[hfuru]

Neither conn_recv_string() nor its callers free *newbuffer on error.

The promotion rules break negative 'res' for 64-bit Py_ssize_t
in the (ulength <= buflength) branch:
res = -1 ==> (UINT32)-1 ==> Py_ssize_t 0xffffffff instead of -1.

While I'm writing: The _conn_recvall() calls can be factored out
of the if().

This patch applies to both 3.2a3 and 2.7. However, the patched failure
cases are untested: I do not know how to test them. It passes make test.

[python-dev]

New changeset 60a7b704de5c by Richard Oudkerk in branch '2.7':
Issue bpo-10133: Make multiprocessing deallocate buffer if socket read fails.
http://hg.python.org/cpython/rev/60a7b704de5c

New changeset 5643697070c0 by Richard Oudkerk in branch '3.2':
Issue bpo-10133: Make multiprocessing deallocate buffer if socket read fails.
http://hg.python.org/cpython/rev/5643697070c0

[sbt]

Thanks for the patch, I have applied it. (I don't think there was a problem with the promotion rules because res was a never converted to UINT32.)

[hfuru]

Richard Oudkerk <report@bugs.python.org> wrote:

Thanks for the patch, I have applied it. (I don't think there was a
problem with the promotion rules because res was a never converted to
UINT32.)

True now that res is a Py_ssize_t. It was int when I wrote the patch.

Hallvard